abstract
stringlengths 383
2.62k
| TLDR
stringlengths 4
250
|
|---|---|
Molecular dynamics (MD) simulation techniques are widely used for various natural science applications. Increasingly, machine learning (ML) force field (FF) models begin to replace ab-initio simulations by predicting forces directly from atomic structures. Despite significant progress in this area, such techniques are primarily benchmarked by their force/energy prediction errors, even though the practical use case would be to produce realistic MD trajectories. We aim to fill this gap by introducing a novel benchmark suite for ML MD simulation. We curate representative MD systems, including water, organic molecules, peptide, and materials, and design evaluation metrics corresponding to the scientific objectives of respective systems. We benchmark a collection of state-of-the-art (SOTA) ML FF models and illustrate, in particular, how the commonly benchmarked force accuracy is not well aligned with relevant simulation metrics. We demonstrate when and how selected SOTA methods fail, along with offering directions for further improvement. Specifically, we identify stability as a key metric for ML models to improve. Our benchmark suite comes with a comprehensive open source codebase for training and simulation with ML FFs to facilitate further work. | We benchmark machine learning force fields with novel datasets and metrics based on molecular simulations and reveal insights into how they should be evaluated and further improved. |
Dynamic benchmarks interweave model fitting and data collection in an attempt to mitigate the limitations of static benchmarks. In contrast to an extensive theoretical and empirical study of the static setting, the dynamic counterpart lags behind due to limited empirical studies and no apparent theoretical foundation to date. Responding to this deficit, we initiate a theoretical study of dynamic benchmarking. We examine two realizations, one capturing current practice and the other modeling more complex settings. In the first model, where data collection and model fitting alternate sequentially, we prove that model performance improves initially but can stall after only three rounds. Label noise arising from, for instance, annotator disagreement leads to even stronger negative results. Our second model generalizes the first to the case where data collection and model fitting have a hierarchical dependency structure. We show that this design guarantees strictly more progress than the first, albeit at a significant increase in complexity. We support our theoretical analysis by simulating dynamic benchmarks on two popular datasets. These results illuminate the benefits and practical limitations of dynamic benchmarking, providing both a theoretical foundation and a causal explanation for observed bottlenecks in empirical work. | We propose a formal model of dynamic benchmarks illuminating their benefits and limitations. |
We propose DITTO, a fully offline approach to imitation learning which addresses the problem of covariate shift without access to an oracle or any additional online interactions. By unrolling agent policies in the latent space of a learned world model and penalizing drift from expert demonstrations, we can use online reinforcement learning algorithms to learn policies which solve the imitation objective, without access to the underlying environment or reward function. Decoupling policy and world model learning lets us leverage datasets of any quality to learn latent representations which provide a natural reward signal for imitation learning, avoiding the need for complex adversarial or sparse imitation-inducing rewards. Compared to competitive baselines, our method achieves state-of-the-art performance in a variety of challenging environments from pixel observations alone. | Completely offline imitation learning with world models, using RL on a latent matching objective in the model. |
Coordinating with previously unknown teammates without joint learning is a crucial need for real-world multi-agent applications, such as human-AI interaction. An active research topic on this problem is ad hoc teamwork, which improves agents' coordination ability in zero-shot settings. However, previous works can only solve the problem of a single agent's coordination with different teams, which is not in line with arbitrary group-to-group coordination in complex multi-agent scenarios. Moreover, they commonly suffer from limited adaptation ability within an episode in a zero-shot setting. To address these problems, we introduce the Coordination Scheme Probing (CSP) approach that applies a disentangled scheme probing module to represent and classify the newly arrived teammates beforehand with limited pre-collected episodic data and makes multi-agent control accordingly. To achieve generalization, CSP learns a meta-policy with multiple sub-policies that follow distinguished coordination schemes in an end-to-end fashion and automatically reuses it to coordinate with unseen teammates. Empirically, we show that the proposed method achieves remarkable performance compared to existing ad hoc teamwork and policy generalization methods in various multi-agent cooperative scenarios. | A few-shot MARL approach for improving coordination ability with diverse teammates |
Unsupervised visual representation learning offers the opportunity to leverage large corpora of unlabeled trajectories to form useful visual representations, which can benefit the training of reinforcement learning (RL) algorithms. However, evaluating the fitness of such representations requires training RL algorithms which is both computationally intensive and has high variance outcomes. To alleviate this issue, we design an evaluation protocol for unsupervised RL representations with lower variance and up to 600x lower computational cost. Inspired by the vision community, we propose two linear probing tasks: predicting the reward observed in a given state, and predicting the action of an expert in a given state. These two tasks are generally applicable to many RL domains, and we show through rigorous experimentation that they correlate strongly with the actual downstream control performance on the Atari100k Benchmark. This provides a better method for exploring the space of pretraining algorithms without the need of running RL evaluations for every setting. Leveraging this framework, we further improve existing self-supervised learning (SSL) recipes for RL, highlighting the importance of the forward model, the size of the visual backbone, and the precise formulation of the unsupervised objective. | Our paper proposes linear reward probing as an efficient method to evaluate the quality of pretrained representations in the RL setting, and demonstrates its positive correlation with downstream RL performance. |
Interpretability in deep learning is one of the largest obstacles to its more widespread adoption in critical applications. A variety of methods have been introduced to understand and explain decisions made by Deep Models. A class of these methods highlights which features are most influential to model predictions. These methods have some key weaknesses. First, most of these methods are applicable only to the atomic elements that make up raw inputs to the model (e.g., pixels or words). Second, these methods generally don't distinguish between the importance of features individually versus due to interactions with other features. As a result, it is difficult to explore high level questions about how models use features. We tackle these issues by proposing Sample-Based Semantic Analysis (SBSA). We use Sobol sensitivity analysis as our sample-based method. Sobol-SBSA allows us to quantify the importance of semantic combinations of raw inputs and highlight the extent to which these features are important individually as opposed to due to interactions with other features. We demonstrate the ability of Sobol-SBSA to answer a richer class of questions about the behavior of Deep Learning models by exploring how CNN models from AlexNet to DenseNet use regions when classifying images. We present two key findings. 1) The architectural improvements from AlexNet to DenseNet manifested themselves in CNN models utilizing greater levels of region interactions for predictions. 2) Adversarially robust CNNs resist exploiting spurious correlations in ImageNet data by forcing these architectures to rely less on region-to-region interaction. Our proposed method is generalizable to a wide variety of network and input types and can help provide greater clarity about model decisions. | This paper introduces a semantic interpretability framework that is used to understand how CNN models and their robust counterparts manipulate image regions. |
Linear time-invariant state space models (SSM) are a classical model from engineering and statistics, that have recently been shown to be very promising in machine learning through the Structured State Space sequence model (S4). A core component of S4 involves initializing the SSM state matrix to a particular matrix called a HiPPO matrix, which was empirically important for S4's ability to handle long sequences. However, the specific matrix that S4 uses was actually derived in previous work for a particular *time-varying* dynamical system, and the use of this matrix as a *time-invariant* SSM had no known mathematical interpretation. Consequently, the theoretical mechanism by which S4 models long-range dependencies actually remains unexplained. We derive a more general and intuitive formulation of the HiPPO framework, which provides a simple mathematical interpretation of S4 as a decomposition onto exponentially-warped Legendre polynomials, explaining its ability to capture long dependencies. Our generalization introduces a theoretically rich class of SSMs that also lets us derive more intuitive S4 variants for other bases such as the Fourier basis, and explains other aspects of training S4, such as how to initialize the important timescale parameter. These insights improve S4's performance to 86% on the Long Range Arena benchmark, with 96% on the most difficult Path-X task.
| We develop a new theoretical interpretation of S4 and generalize it to other basis functions |
Real-world graphs such as social networks, communication networks, and rating networks are constantly evolving over time. Many architectures have been developed to learn effective node representations using both graph structure and its dynamics. While the robustness of static graph models is well-studied, the vulnerability of the dynamic graph models to adversarial attacks is underexplored. In this work, we design a novel adversarial attack on discrete-time dynamic graph models where we desire to perturb the input graph sequence in a manner that preserves the temporal dynamics of the graph. To this end, we motivate a novel Temporal Dynamics-Aware Perturbation (TDAP) constraint, which ensures that perturbations introduced at each time step are restricted to only a small fraction of the number of changes in the graph since the previous time step. We present a theoretically-grounded Projected Gradient Descent approach for dynamic graphs to find the effective perturbations under the TDAP constraint. Experiments on two tasks — dynamic link prediction and node classification, show that our approach is up to 4x more effective than the baseline methods for attacking these models. We also consider the practical online setting where graph snapshots become available in real-time and extend our attack approach to use Online Gradient Descent for performing attacks under the TDAP constraint. In this more challenging setting, we demonstrate that our method achieves upto 5x superior performance when compared to representative baselines. | Introduces a novel constraint to attack dynamic graph models while preserving the original graph evolution and presents an effective approach to find such attacks |
Certified defenses against small-norm adversarial examples have received growing attention in recent years; though certified accuracies of state-of-the-art methods remain far below their non-robust counterparts, despite the fact that benchmark datasets have been shown to be well-separated at far larger radii than the literature generally attempts to certify. In this work, we offer insights that identify potential factors in this performance gap. Specifically, our analysis reveals that piecewise linearity imposes fundamental limitations on the tightness of leading certification techniques. These limitations are felt in practical terms as a greater need for capacity in models hoped to be certified efficiently. Moreover, this is _in addition_ to the capacity necessary to learn a robust boundary, studied in prior work. However, we argue that addressing the limitations of piecewise linearity through scaling up model capacity may give rise to potential difficulties---particularly regarding robust generalization---therefore, we conclude by suggesting that developing _smooth_ activation functions may be the way forward for advancing the performance of certified neural networks. | We show that piecewise linearity imposes fundamental limitations for efficient robustness certification, e.g., Lipschitz-based certification; this imposes additional capacity requirements on networks that must be certified by such techniques. |
Denoising diffusion models have spurred significant gains in density modeling and image generation, precipitating an industrial revolution in text-guided AI art generation. We introduce a new mathematical foundation for diffusion models inspired by classic results in information theory that connect Information with Minimum Mean Square Error regression, the so-called I-MMSE relations. We generalize the I-MMSE relations to \emph{exactly} relate the data distribution to an optimal denoising regression problem, leading to an elegant refinement of existing diffusion bounds. This new insight leads to several improvements for probability distribution estimation, including a theoretical justification for diffusion model ensembling. Remarkably, our framework shows how continuous and discrete probabilities can be learned with the same regression objective, avoiding domain-specific generative models used in variational methods. | A new information-theoretic foundation for diffusion models leads to simpler and more computationally efficient density modeling. |
Machine learning methods are getting increasingly better at making predictions, but at the same time they are also becoming more complicated and less transparent. As a result, explainers are often relied on to provide interpretability to these \textit{black-box} prediction models. As crucial diagnostics tools, it is important that these explainers themselves are reliable. In this paper we focus on one particular aspect of reliability, namely that an explainer should give similar explanations for similar data inputs. We formalize this notion by introducing and defining \textit{explainer astuteness}, analogous to astuteness of classifiers. Our formalism is inspired by the concept of \textit{probabilistic Lipschitzness}, which captures the probability of local smoothness of a function. For a variety of explainers (e.g., SHAP, RISE, CXPlain), we provide lower bound guarantees on the astuteness of these explainers given the Lipschitzness of the prediction function. These theoretical results imply that locally smooth prediction functions lend themselves to locally robust explanations. We evaluate these results empirically on simulated as well as real datasets. | Theoretical work in support of the intuition that robust classifiers lend themselves to robust explainers |
Typical continual learning setup assumes that the dataset is split into multiple discrete tasks. We argue that it is less realistic as the streamed data would have no notion of task boundary in real-world data. Here, we take a step forward to investigate more realistic online continual learning – learning continuously changing data distribution without explicit task boundary, which we call boundary-free setup. As there is no clear boundary of tasks, it is not obvious when and what information in the past to be preserved as a better remedy for the stability-plasticity dilemma. To this end, we propose a scheduled transfer of previously learned knowledge. We further propose a data-driven balancing between the knowledge in the past and the present in learning objective. Moreover, since it is not straight-forward to use the previously proposed forgetting measure without task boundaries, we further propose a novel forgetting measure based on information theory that can capture forgetting. We empirically evaluate our method on a Gaussian data stream, its periodic extension, which assumes periodic data distribution frequently observed in real-life data, as well as the conventional disjoint task-split. Our method outperforms prior arts by large margins in various setups, using four popular benchmark datasets – CIFAR-10, CIFAR-100, TinyImageNet and ImageNet. | We propose a new continual learning setup without explicit task boundary and a method to address it. |
Designing a neural network architecture for molecular representation is crucial for AI-driven drug discovery and molecule design. In this work, we propose a new framework for molecular representation learning. Our contribution is threefold: (a) demonstrating the usefulness of incorporating substructures to node-wise features from molecules, (b) designing two branch networks consisting of a transformer and a graph neural network so that the networks fused with asymmetric attention, and (c) not requiring heuristic features and computationally-expensive information from molecules. Using 1.8 million molecules collected from ChEMBL and PubChem database, we pretrain our network to learn a general representation of molecules with minimal supervision. The experimental results show that our pretrained network achieves competitive performance on 11 downstream tasks for molecular property prediction. | This paper proposes a novel network that utilizes molecular substructure along with local atom embedding. |
We propose Waveformer that learns attention mechanism in the wavelet coefficient space, requires only linear time complexity, and enjoys universal approximating power. Specifically, we first apply forward wavelet transform to project the input sequences to multi-resolution orthogonal wavelet bases, then conduct nonlinear transformations (in this case, a random feature kernel) in the wavelet coefficient space, and finally reconstruct the representation in input space via backward wavelet transform. We note that other non-linear transformations may be used, hence we name the learning paradigm Wavelet transformatIon for Sequence lEarning (WISE). We emphasize the importance of backward reconstruction in the WISE paradigm — without it, one would be mixing information from both the input space and coefficient space through skip connections, which shall not be considered as mathematically sound. Compared with Fourier transform in recent works, wavelet transform is more efficient in time complexity and better captures local and positional information; we further support this through our ablation studies. Extensive experiments on seven long-range understanding datasets from the Long Range Arena benchmark and code understanding tasks demonstrate that (1) Waveformer achieves competitive and even better accuracy than a number of state-of-the-art Transformer variants and (2) WISE can boost accuracies of various attention approximation methods without increasing the time complexity. These together showcase the superiority of learning attention in a wavelet coefficient space over the input space. | We propose a model with linear complexity achieving SOTA results on a set of long-range tasks, under a new paradigm to learn attention in wavelet space which boosts accuracies of various attention methods without increasing time complexity. |
Adversarial attacks in NLP challenge the way we look at language models. The goal of this kind of adversarial attack is to modify the input text to fool a classifier while maintaining the original meaning of the text. Although most existing adversarial attacks claim to fulfill the constraint of semantics preservation, careful scrutiny shows otherwise. We show that the problem lies in the text encoders used to determine the similarity of adversarial examples, specifically in the way they are trained. Unsupervised training methods make these encoders more susceptible to problems with antonym recognition. To overcome this, we introduce a simple, fully supervised sentence embedding technique called Semantics-Preserving-Encoder (SPE). The results show that our solution minimizes the variation in the meaning of the adversarial examples generated. It also significantly improves the overall quality of adversarial examples, as confirmed by human evaluators. Furthermore, it can be used as a component in any existing attack to speed up its execution while maintaining similar attack success. | We propose a novel sentence encoder that improves the quality of textual adversarial attacks. |
Neural Architecture Search (NAS) aims to search for the best network in the pre-defined search space. However, much work focuses on the search strategy but little on the architecture selection process. Despite the fact that the weight-sharing based NAS has promoted the search efficiency, we notice that the architecture selection is quite unstable or circuitous. For instance, the differentiable NAS may derive the suboptimal architecture due to the performance collapse caused by bi-level optimization, or the One-shot NAS requires sampling and evaluating a large number of candidate structures. Recently, the self-attention mechanism achieves better performance in terms of the long-range modeling capabilities. Considering that different operations are widely distributed in the search space, we suggest leveraging the self-attention mechanism to extract the relationship among them and to determine which operation is superior to others. Therefore, we integrate Lite-Transformer into NAS for architecture selection. Specifically, we regard the feature map of each candidate operation as distinct patches and feed them into the Lite-Transformer module along with an additional Indicator Token (called IT). The cross attention among various operations can be extracted by the self-attention mechanism, and the importance of each candidate operation is then shown by the softmax result between the query of indicator token (IT) and other values of operational tokens. We experimentally demonstrate that our framework can select the truly representative architecture in different search spaces and achieves 2.39% test error on CIFAR-10 in DARTS search space, and 24.1% test error on ImageNet in the ProxylessNAS search space, as well as the stable and better performance in NAS-Bench-201 search space and S1-S4 search spaces, outperforming state-of-the-art NAS methods. | This paper proposes to integrate Lite-Transformer into NAS for architecture selection, and introduces an additional indicator token (IT) to reflect the importance of each candidate operation. |
The success of the Adam optimizer on a wide array of architectures has made it the default in settings where stochastic gradient descent (SGD) performs poorly. However, our theoretical understanding of this discrepancy is lagging, preventing the development of significant improvements on either algorithm. Recent work advances the hypothesis that Adam and other heuristics like gradient clipping outperform SGD on language tasks because the distribution of the error induced by sampling has heavy tails. This suggests that Adam outperform SGD because it uses a more robust gradient estimate. We evaluate this hypothesis by varying the batch size, up to the entire dataset, to control for stochasticity. We present evidence that stochasticity and heavy-tailed noise are not major factors in the performance gap between SGD and Adam. Rather, Adam performs better as the batch size increases, while SGD is less effective at taking advantage of the reduction in noise. This raises the question as to why Adam outperforms SGD in the full-batch setting. Through further investigation of simpler variants of SGD, we find that the behavior of Adam with large batches is similar to sign descent with momentum. | A hypothesis for Adam's success is that it handles heavy-tailed noise better than SGD, but it works even better without noise; with big batch sizes, it performs very similarly to sign descent, which might help explain why it works. |
Existing CNN architectures may achieve efficiency in either one or two dimensions: FLOPs, depth, accuracy, representation power, latency but not in all. In this work, we present a pragmatically designed novel CNN architecture “CoMNet” which offers multi-dimensional efficiency at once such as: simple yet accurate, lower latency and FLOPs, high representation power in limited parameters, low memory consumption, negligible branching, smaller depths, and only a few design hyperparameters. The key to achieve the multi-dimensional efficiency is our use of biological underpinnings into CoMNet which is primarily the organization of cortical modules in the visual cortex. To realize CoMNet, a few concepts from well understood CNN designs are directly inherited such as residual learning. Our solid experimental evaluations demonstrate superiority of CoMNet over many state-of-the-art industry and academia dominant architectures such as ResNet, RepVGG etc. For instance, CoMNet supersedes ResNet-50 on ImageNet while being 50% shallower, 22% lesser parameters, 25% lower FLOPs and latency, and in 16% lesser training epochs. Code will be opensourced post the reviews. | A novel CNN architecture leveraging biological structures in visual cortex to cater real-time applications with low latency, smaller depths |
Neural Processes (NPs) are popular methods in meta-learning that can estimate predictive uncertainty on target datapoints by conditioning on a context dataset. Previous state-of-the-art method Transformer Neural Processes (TNPs) achieve strong performance but require quadratic computation with respect to the number of context datapoints, significantly limiting its scalability. Conversely, existing sub-quadratic NP variants perform significantly worse than that of TNPs. Tackling this issue, we propose Latent Bottlenecked Attentive Neural Processes (LBANPs), a new computationally efficient sub-quadratic NP variant, that has a querying computational complexity independent of the number of context datapoints. The model encodes the context dataset into a constant number of latent vectors on which self-attention is performed. When making predictions, the model retrieves higher-order information from the context dataset via multiple cross-attention mechanisms on the latent vectors. We empirically show that LBANPs achieve results competitive with the state-of-the-art on meta-regression, image completion, and contextual multi-armed bandits. We demonstrate that LBANPs can trade-off the computational cost and performance according to the number of latent vectors. Finally, we show LBANPs can scale beyond existing attention-based NP variants to larger dataset settings. | In this work, we propose Latent Bottlenecked Attentive Neural Processes (LBANPs), a new computationally efficient sub-quadratic NP variant, that has a querying computational complexity independent of the number of context datapoints. |
In this work, we propose an alternative to the Marginal Likelihood (MaL) objective for training latent variable models, Complete Latent Likelihood (CoLLike). We analyze the objectives from the perspective of matching joint distributions. We show that MaL corresponds to a particular $KL$ divergence between some target \emph{joint} distribution and the model joint. Furthermore, the properties of the target joint explain such major malfunctions of MaL as uninformative latents (posterior collapse) and high deviation of the aggregated posterior from the prior. In CoLLike approach, we use a sample from the prior to construct a family of target joint distributions, which properties prevent these drawbacks. We utilize the complete likelihood both to choose the target from this family and to learn the model. We confirm our analysis by experiments with expressive low-dimensional latent variable models, which also indicate that it is possible to achieve high accuracy unsupervised classification using CoLLike objective. | Use sample from the prior to construct a family informative distribution and use complete likelihood to both the target from the family and tune the model. |
The key problem in multivariate time series (MTS) analysis and forecasting aims to disclose the underlying couplings between variables that drive the co-movements. Considerable recent successful MTS methods are built with graph neural networks (GNNs) due to their essential capacity for relational modeling. However, previous work often used a static graph structure of time-series variables for modeling MTS failing to capture their ever-changing correlations over time. To this end, a fully-connected supra-graph connecting any two variables at any two timestamps is adaptively learned to capture the high-resolution variable dependencies via an efficient graph convolutional network. Specifically, we construct the Edge-Varying Fourier Graph Networks (EV-FGN) equipped with Fourier Graph Shift Operator (FGSO) which efficiently performs graph convolution in the frequency domain. As a result, a high-efficiency scale-free parameter learning scheme is derived for MTS analysis and forecasting according to the convolution theorem. Extensive experiments show that EV-FGN outperforms state-of-the-art methods on seven real-world MTS datasets. | We make the first attempt to design a complex-valued feed-forward network in the Fourier space for efficiently computing multi-layer graph convolutions with defined Fourier graph shift operators. |
In many real world settings binary classification decisions are made based on limited data in near real-time, e.g. when assessing a loan application. We focus on a class of these problems that share a common feature that the true label is only observed when a data point is assigned a positive label by a learner, e.g. we only learn of an outcome of \emph{accepted} loan applications. In this setting, sometimes referred to as the Bank Loan Problem (BLP) in the literature, the labelled training set suffers from accumulating bias since it is created by learners past decisions.
Prior work mitigates the consequences of this bias by injecting optimism into the model to allow the learner to correct self-reinforcing false rejections. This reduces long term regret but comes at the cost of a higher false acceptance rate.
We introduce \emph{adversarial optimism} (AdOpt) to directly address the bias in the training set using \emph{adversarial domain adaptation}. The goal of AdOpt is to learn an unbiased but informative representation of past data, by reducing the distributional shift between the set of \textit{accepted} data points and all data points seen thus far. AdOpt integrates classification made using this debiased representation of the data with the recently proposed \emph{pseudo-label optimism}(PLOT) method to increase the rate of correct decisions at every timestep.
AdOpt significantly exceeds state-of-the-art performance on a set of challenging BLP benchmark problems. | We use adversarial domain adaptation to combat accumulating bias for a class of online learning problems. |
We study the infinite horizon average reward constrained Markov Decision Process (CMDP). In contrast to existing works on model-based, finite state space, we consider the model-free linear CMDP setup. We first propose a computationally inefficient algorithm and show that $\tilde{\mathcal{O}}(\sqrt{d^3T})$ regret and constraint violation can be achieved, in which $T$ is the number of interactions, and $d$ is the dimension of the feature mapping. We also propose an efficient variant based on the primal-dual adaptation of the LSVI-UCB algorithm and show that $\tilde{\mathcal{O}}((dT)^{3/4})$ regret and constraint violation can be achieved.
This improves the known regret bound of $\tilde{\mathcal{O}}(T^{5/6})$ for the finite state-space model-free constrained RL which was obtained under a stronger assumption compared to ours. We also develop an efficient policy-based algorithm via novel adaptation of the MDP-EXP2 algorithm to our primal-dual set up with $\tilde{\mathcal{O}}(\sqrt{T})$ regret and even zero constraint violation bound under a stronger set of assumptions. | We provide the first sub-linear regret and sub-linear constraint violation for constrained MDP for linear function approximation using model-free RL algorithm |
In cooperative multi-agent reinforcement learning (MARL), combining value decomposition with actor-critic enables agents to learn stochastic policies, which are more suitable for the partially observable environment. Given the goal of learning local policies that enable decentralized execution, agents are commonly assumed to be independent of each other, even in centralized training. However, such an assumption may prohibit agents from learning the optimal joint policy. To address this problem, we explicitly take the dependency among agents into centralized training. Although this leads to the optimal joint policy, it may not be factorized for decentralized execution. Nevertheless, we theoretically show that from such a joint policy, we can always derive another joint policy that achieves the same optimality but can be factorized for decentralized execution. To this end, we propose multi-agent conditional policy factorization (MACPF), which takes more centralized training but still enables decentralized execution. We empirically verify MACPF in various cooperative MARL tasks and demonstrate that MACPF achieves better performance or faster convergence than baselines. Our code is available at https://github.com/PKU-RL/FOP-DMAC-MACPF. | We propose a novel multi-agent reinforcement learning algorithm where dependency among agents is explicitly considered |
While standard bandit algorithms sometimes incur high regret, their performance can be greatly improved by "warm starting" with historical data. Unfortunately, how best to incorporate historical data is unclear: naively initializing reward estimates using all historical samples can suffer from spurious data and imbalanced data coverage, leading to computational and storage issues - particularly in continuous action spaces. We address these two challenges by proposing Artificial Replay, a meta-algorithm for incorporating historical data into any arbitrary base bandit algorithm. Artificial Replayuses only a subset of the historical data as needed to reduce computation and storage. We show that for a broad class of base algorithms that satisfy independence of irrelevant data (IIData), a novel property that we introduce, our method achieves equal regret as a full warm-start approach while potentially using only a fraction of historical data. We complement these theoretical results with a case study of $K$-armed and continuous combinatorial bandit algorithms, including on a green security domain using real poaching data, to show the practical benefits of Artificial Replayin achieving optimal regret alongside low computational and storage costs. | We propose a meta-algorithm for multi-armed bandits that most efficiently uses historical data, overcoming challenges of spurious data and imbalanced data coverage. |
Noisy labels are inevitably presented in real-world datasets due to labeling error or visual content ambiguity. Existing methods generally approach the task of noisy label learning (NLL) by either properly regularizing the model, or reweighting clean/noisy labeled samples. While self-supervised learning (SSL) has been applied to pre-train deep neural networks without label supervision, downstream tasks like image classification still require clean labeled data. And, most SSL strategies are performed at the instance level, regardless of the correctness of its label. In this paper, we propose set-level self-supervised learning (SLSSL), which performs SSL at mini-batch levels with observed noisy labels. By corrupting the labels of each training mini-batch, our SLSSL enforces the model to exhibit sufficient robustness. Moreover, the proposed SLSSL can also be utilized for sample reweighting technique. As a result, the proposed learning scheme can be applied as an expectation-maximization (EM) algorithm during model training. Extensive experiments on synthetic and real-world noisy label data confirm the effectiveness of our framework. | This paper proposes a set-level self-supervised learning techniques on each training mini-batch to tackle noisy-label learning problems. |
Decision Transformers (DT) have demonstrated strong performances in offline reinforcement learning settings, but quickly adapting to unseen novel tasks remains challenging. To address this challenge, we propose a new framework, called Hyper-Decision Transformer (HDT), that can generalize to novel tasks from a handful of demonstrations in a data- and parameter-efficient manner. To achieve such a goal, we propose to augment the base DT with an adaptation module, whose parameters are initialized by a hyper-network. When encountering unseen tasks, the hyper-network takes a handful of demonstrations as inputs and initializes the adaptation module accordingly. This initialization enables HDT to efficiently adapt to novel tasks by only fine-tuning the adaptation module. We validate HDT's generalization capability on object manipulation tasks. We find that with a single expert demonstration and fine-tuning only 0.5% of DT parameters, HDT adapts faster to unseen tasks than fine-tuning the whole DT model. Finally, we explore a more challenging setting where expert actions are not available, and we show that HDT outperforms state-of-the-art baselines in terms of task success rates by a large margin. Demos are available on our project page: https://sites.google.com/view/hdtforiclr2023/home. | We propose Hyper-Decision Transformer (HDT), a transformer-based model which generalizes to novel unseen tasks maintaining strong data and parameter efficiency. |
Chain-of-thought prompting combined with pretrained large language models has achieved encouraging results on complex reasoning tasks. In this paper, we propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting. It first samples a diverse set of reasoning paths instead of only taking the greedy one, and then selects the most consistent answer by marginalizing out all possible reasoning paths. Self-consistency leverages the intuition that a complex reasoning problem typically admits multiple different ways of thinking leading to its unique correct answer. Our extensive empirical evaluation shows that self-consistency boosts the performance of chain-of-thought prompting with a striking margin on a range of popular arithmetic and commonsense reasoning benchmarks, including GSM8K (+17.9%), SVAMP (+11.0%), AQuA (+12.2%), StrategyQA (+6.4%) and ARC-challenge (+3.9%). | We propose a new decoding strategy, self-consistency, that greatly improves chain-of-thought prompting |
Incorporated with the recent advances in deep learning, deep reinforcement learning (DRL) has achieved tremendous success in empirical study. However, analyzing DRL is still challenging due to the complexity of the neural network class. In this paper, we address such a challenge by analyzing the Markov decision process (MDP) with neural dynamics, which covers several existing models as special cases, including the kernelized nonlinear regulator (KNR) model and the linear MDP. We propose a novel algorithm that designs exploration incentives via learnable representations of the dynamics model by embedding the neural dynamics into a kernel space induced by the system noise. We further establish an upper bound on the sample complexity of the algorithm, which demonstrates the sample efficiency of the algorithm. We highlight that, unlike previous analyses of RL algorithms with function approximation, our bound on the sample complexity does not depend on the Eluder dimension of the neural network class, which is known to be exponentially large (Dong et al., 2021). | We identify a class of Markov decision processes with neural network parameterization and propose an oracle-efficient algorithm whose sample complexity does not depend on the Eluder dimension of the NN class. |
Serving a GNN model online is challenging --- in many applications when testing nodes are connected to training nodes, one has to propagate information from training nodes to testing nodes to achieve the best performance, and storing the whole training set (including training graph and node features) during inference stage is prohibitive for large-scale problems. In this paper, we study graph compression to reduce the storage requirement for GNN in serving. Given a GNN model to be served, we propose to construct a compressed graph with a smaller number of nodes. In serving time, one just needs to replace the original training set graph by this compressed graph, without the need of changing the actual GNN model and the forward pass. We carefully analyze the error in the forward pass and derive simple ways to construct the compressed graph to minimize the approximation error. Experimental results on semi-supervised node classification demonstrate that the proposed method can significantly reduce the serving space requirement for GNN inference. | Compressing the graph for graph neural networks inference |
Implicit graph neural networks (IGNNs) that solve a fixed-point equilibrium equation for representation learning can learn the long-range dependencies (LRD) in the underlying graphs and show remarkable performance for various graph learning tasks. However, the expressivity of IGNNs is limited by the constraints for their well-posedness guarantee. Moreover, when IGNNs become effective for learning LRD, their eigenvalues converge to the value that slows down the convergence, and their performance is unstable across different tasks. In this paper, we provide a new well-posedness condition of IGNNs leveraging monotone operator theory. The new well-posedness characterization informs us to design effective parameterizations to improve the accuracy, efficiency, and stability of IGNNs. Leveraging accelerated operator splitting schemes and graph diffusion convolution, we design efficient and flexible implementations of monotone operator IGNNs that are significantly faster and more accurate than existing IGNNs. | We propose stable, efficient, and flexible implicit graph neural networks leveraging monotone operator theory |
Sampling from a target measure whose density is only known up to a normalization constant is a fundamental problem in computational statistics and machine learning. In this paper, we present a new optimization-based method for sampling called mollified interaction energy descent (MIED). MIED minimizes a new class of energies on probability measures called mollified interaction energies (MIEs). These energies rely on mollifier functions---smooth approximations of the Dirac delta originated from PDE theory. We show that as the mollifier approaches the Dirac delta, the MIE converges to the chi-square divergence with respect to the target measure and the gradient flow of the MIE agrees with that of the chi-square divergence. Optimizing this energy with proper discretization yields a practical first-order particle-based algorithm for sampling in both unconstrained and constrained domains. We show experimentally that for unconstrained sampling problems our algorithm performs on par with existing particle-based algorithms like SVGD, while for constrained sampling problems our method readily incorporates constrained optimization techniques to handle more flexible constraints with strong performance compared to alternatives.
| Unconstrained and constrained sampling by minimizing a new class of mollified interaction energies. |
Transformer has achieved remarkable success in language, image, and speech processing. Recently, various efficient attention architectures have been proposed to improve transformer's efficiency while largely preserving its efficacy, especially in modeling long sequences. A widely-used benchmark to test these efficient methods' capability on long-range modeling is Long Range Arena (LRA). However, LRA only focuses on the standard bidirectional (or noncausal) self attention, and completely ignores cross attentions and unidirectional (or causal) attentions, which are equally important to downstream applications. Although designing cross and causal variants of an attention method is straightforward for vanilla attention, it is often challenging for efficient attentions with subquadratic time and memory complexity. In this paper, we propose Comprehensive Attention Benchmark (CAB) under a fine-grained attention taxonomy with four distinguishable attention patterns, namely, noncausal self, causal self, noncausal cross, and causal cross attentions. CAB collects seven real-world tasks from different research areas to evaluate efficient attentions under the four attention patterns. Among these tasks, CAB validates efficient attentions in eight backbone networks to show their generalization across neural architectures. We conduct exhaustive experiments to benchmark the performances of nine widely-used efficient attention architectures designed with different philosophies on CAB. Extensive experimental results also shed light on the fundamental problems of efficient attentions, such as efficiency length against vanilla attention, performance consistency across attention patterns, the benefit of attention mechanisms, and interpolation/extrapolation on long-context language modeling. | We propose Comprehensive Attention Benchmark (CAB) with seven real-world tasks from different research areas to evaluate efficient attentions under four fine-grained attention patterns. |
Recent work has shown that Deep Reinforcement Learning (DRL) is vulnerable to adversarial attacks, in which minor perturbations of input signals cause agents to behave inappropriately and unexpectedly. Humans, on the other hand, appear robust to these particular sorts of input variations. We posit that this part of robustness stems from accumulated knowledge about the world.
In this work, we propose to leverage prior knowledge to defend against adversarial attacks in RL settings using a framework we call Knowledge-based Policy Recycling (KPR). Different from previous defense methods such as adversarial training and robust learning, KPR incorporates domain knowledge over a set of auxiliary tasks policies and learns relations among them from interactions with the environment via a Graph Neural Network (GNN). KPR can use any relevant policy as an auxiliary policy and, importantly, does not assume access or information regarding the adversarial attack. Empirically, KPR results in policies that are more robust to various adversarial attacks in Atari games and a simulated Robot Foodcourt environment. | Incorporating domain-knowledge over auxiliary tasks enhances deep reinforcement policy robustness against adversarial attacks in both Atari games and a high dimensional Robot Food Court environment. |
Optimizing multiple competing objectives is a common problem across science and industry. The inherent inextricable trade-off between those objectives leads one to the task of exploring their Pareto front. A meaningful quantity for the purpose of the latter is the hypervolume indicator, which is used in Bayesian Optimization (BO) and Evolutionary Algorithms (EAs). However, the computational complexity for the calculation of the hypervolume scales unfavorably with increasing number of objectives and data points, which restricts its use in those common multi-objective optimization frameworks.
To overcome these restrictions, previous work has focused on approximating the hypervolume using deep learning. In this work, we propose a novel deep learning architecture to approximate the hypervolume function, which we call DeepHV. For better sample efficiency and generalization, we exploit the fact that the hypervolume is scale equivariant in each of the objectives as well as permutation invariant w.r.t. both the objectives and the samples, by using a deep neural network that is equivariant w.r.t. the combined group of scalings and permutations. We show through an ablation study that including these symmetries leads to significantly improved model accuracy.
We evaluate our method against exact, and approximate hypervolume methods in terms of accuracy, computation time, and generalization. We also apply and compare our methods to state-of-the-art multi-objective BO methods and EAs on a range of synthetic and real-world benchmark test cases. The results show that our methods are promising for such multi-objective optimization tasks. | Hypervolume approximation using permutation invariant, scaling equivariant neural network |
We propose a synthetic reasoning task, LEGO (Learning Equality and Group Operations), that encapsulates the problem of following a chain of reasoning, and we study how the Transformer architectures learn this task. We pay special attention to data effects such as pretraining (on seemingly unrelated NLP tasks) and dataset composition (e.g., differing chain length at training and test time), as well as architectural variants such as weight-tied layers or adding convolutional components. We study how the trained models eventually succeed at the task, and in particular, we are able to understand (to some extent) some of the attention heads as well as how the information flows in the network. Based on these observations we propose a hypothesis that here pretraining helps for LEGO tasks due to certain structured attention patterns, and we experimentally verify this hypothesis. We also observe that in some data regimes the trained transformer finds ``shortcut" solutions to follow the chain of reasoning, which impedes the model's robustness, and moreover we propose ways to prevent it. Motivated by our findings on structured attention patterns, we propose to replace certain attention heads with hardcoded patterns. This architectural change significantly reduces Flops and maintains or even improves the model's performance at large-scale pretraining. | We propose a synthetic task for logical reasoning on which we study transformer models' intriguing behaviors regarding generalization and the role of pre-training; we gain insights leading to large-scale practical improvements. |
Large language models (LLMs) have a substantial capacity for high-level analogical reasoning: reproducing patterns in linear text that occur in their training data (zero-shot evaluation) or in the provided context (few-shot in-context learning). However, recent studies show that even the largest LLMs fail in scenarios that require reasoning over multiple objects or facts or making sequences of logical deductions. We propose a two-stage probabilistic inference paradigm, ThinkSum, that reasons over sets of objects or facts in a structured manner. In the first stage (Think -- 'fast' retrieval of associations), a LLM is queried in parallel over a set of phrases extracted from the prompt or an auxiliary model call. In the second stage (Sum -- 'slow' probabilistic inference or reasoning), the results of these queries are aggregated to make the final prediction. We demonstrate the advantages of ThinkSum on the BIG-bench suite of evaluation tasks, achieving improvements over the state of the art using GPT-family models on ten difficult tasks, often with far smaller model variants. We compare and contrast ThinkSum with other proposed modifications to direct prompting of LLMs, such as variants of chain-of-thought prompting. We argue that because the probabilistic inference in ThinkSum is performed outside of calls to the LLM, ThinkSum is less sensitive to prompt design, yields more interpretable predictions, and can be flexibly combined with latent variable models to extract structured knowledge from LLMs. | A wise System 2 for large language models: Think (parallel model call) + Sum (aggregate results to make a prediction). |
Post-hoc explanations of deep neural networks improve human understanding on the learned representations, decision-making process and uncertainty of the model with faithfulness. Explaining deep convolutional neural networks(DCNN) is especially challenging, due to the high dimensionality of deep features and the complexity of model inference. Most post-hoc explaining methods serve a single form of explanation, restricting the diversity and consistency of the explanation. This paper proposes joint Gaussian mixture model(JGMM), a probabilistic model jointly models inter-layer deep features and produces faithful and consistent post-hoc explanations. JGMM explains deep features by Gaussian mixture model and inter-layer deep feature relations by posterior distribution on the latent component variables. JGMM enables a versatile explaining framework that unifies interpretable proxy model, global or local explanatory example generation or mining. Experiments are performed on various DCNN image classifiers in comparison with other explaining methods. It shows that JGMM can efficiently produce versatile, consistent, faithful and understandable explanations. | This paper proposes a GMM-based probablistic model to explain DCNN representations and inference by proxy models and explanatory examples. |
Integer Linear Programming (ILP) provides a viable mechanism to encode explicit and controllable assumptions about explainable multi-hop inference with natural language. However, an ILP formulation is non-differentiable and cannot be integrated into broader deep learning architectures. Recently, Thayaparan et al. (2021a) proposed a novel methodology to integrate ILP with Transformers to achieve end-to-end differentiability for complex multi-hop inference. While this hybrid framework has been demonstrated to deliver better answer and explanation selection than transformer-based and existing ILP solvers, the neuro-symbolic integration still relies on a convex relaxation of the ILP formulation, which can produce sub-optimal solutions. To improve these limitations, we propose Diff-Comb Explainer, a novel neuro-symbolic architecture based on Differentiable BlackBox Combinatorial solvers (DBCS) (Pogančić et al., 2019). Unlike existing differentiable solvers, the presented model does not require the transformation and relaxation of the explicit semantic constraints, allowing for direct and more efficient integration of ILP formulations. Diff-Comb Explainer demonstrates improved accuracy and explainability over non-differentiable solvers, Transformers and existing differentiable constraint-based multi-hop inference frameworks. | The paper presents a novel neuro-symbolic framework that integrates Integer Linear Programming with pre-trained transformers to perform end-to-end explainable multi-hop inference |
We present a smoothly broken power law functional form (referred to by us as a broken neural scaling law (BNSL)) that accurately models and extrapolates the scaling behaviors of deep neural networks (i.e. how the evaluation metric of interest varies as the amount of compute used for training, number of model parameters, training dataset size, or upstream performance varies) for various architectures and for each of various tasks within a large and diverse set of upstream and downstream tasks, in zero-shot, prompted, and fine-tuned settings. This set includes large-scale vision, language, audio, video, diffusion, generative modeling, multimodal learning, contrastive learning, AI alignment, robotics, out-of-distribution (OOD) generalization, continual learning, uncertainty estimation / calibration, out-of-distribution detection, adversarial robustness, molecules, computer programming/coding, math word problems, arithmetic, unsupervised/self-supervised learning, and reinforcement learning (single agent and multi-agent). When compared to other functional forms for neural scaling behavior, this functional form yields extrapolations of scaling behavior that are considerably more accurate on this set. Moreover, this functional form accurately models and extrapolates scaling behavior that other functional forms are incapable of expressing such as the non-monotonic transitions present in the scaling behavior of phenomena such as double descent and the delayed, sharp inflection points present in the scaling behavior of tasks such as arithmetic. Lastly, we use this functional form to glean insights about the limit of the predictability of scaling behavior. See arXiv for longer version of this paper. Code is available at https://github.com/ethancaballero/broken_neural_scaling_laws | We present a functional form that accurately models the scaling behaviors for each task from a very large and diverse set of downstream (and upstream) tasks, even scaling behaviors that were previously believed to be "unpredictable". |
Anomaly detection (AD) plays an important role in numerous applications. In this paper, we focus on two understudied aspects of AD that are critical for integration into real-world applications. First, most AD methods cannot incorporate labeled data that are often available in practice in small quantities and can be crucial to achieve high accuracy. Second, most AD methods are not interpretable, a bottleneck that prevents stakeholders from understanding the reason behind the anomalies. In this paper, we propose a novel AD framework, DIAD, that adapts a white-box model class, Generalized Additive Models, to detect anomalies using a partial identification objective which naturally handles noisy or heterogeneous features. DIAD can incorporate a small amount of labeled data to further boost AD performances in semi-supervised settings. We demonstrate the superiority of DIAD compared to previous work in both unsupervised and semi-supervised settings on multiple datasets. We also present explainability capabilities of DIAD, on its rationale behind predicting certain samples as anomalies. | We developed an interpretable tabular anomaly detection method that allows incorporation of labeled data. |
Federated learning is for training a global model without collecting private local data from clients. As they repeatedly need to upload locally-updated weights or gradients instead, clients require both computation and communication resources enough to participate in learning, but in reality their resources are heterogeneous. To enable resource-constrained clients to train smaller local models, width scaling techniques have been used, which reduces the channels of a global model. Unfortunately, width scaling suffers from heterogeneity of local models when averaging them, leading to a lower accuracy than when simply excluding resource-constrained clients from training. This paper proposes a new approach based on depth scaling called DepthFL. DepthFL defines local models of different depths by pruning the deepest layers off the global model, and allocates them to clients depending on their available resources. Since many clients do not have enough resources to train deep local models, this would make deep layers partially-trained with insufficient data, unlike shallow layers that are fully trained. DepthFL alleviates this problem by mutual self-distillation of knowledge among the classifiers of various depths within a local model. Our experiments show that depth-scaled local models build a global model better than width-scaled ones, and that self-distillation is highly effective in training data-insufficient deep layers. | DepthFL is a new federated learning framework based on depth scaling to tackle system heterogeneity. |
Reward design in reinforcement learning (RL) is challenging since specifying human notions of desired behavior may be difficult via reward functions or require many expert demonstrations. Can we instead cheaply design rewards using a natural language interface? This paper explores how to simplify reward design by using a large language model (LLM) such as GPT-3 as a proxy reward function, where the user provides a textual prompt containing a few examples (few-shot) or a description (zero-shot) of desired behavior. Our approach leverages this proxy reward function in an RL framework. Specifically, users specify a prompt once at the beginning of training. During training, the LLM evaluates an RL agent's behavior against the desired behavior described by the prompt and outputs a corresponding reward signal. The RL agent then uses this reward to update its behavior. We evaluate whether our approach can train agents aligned with user objectives in the Ultimatum Game, matrix games, and the DealOrNoDeal negotiation task. In all three tasks, we show that RL agents trained with our framework are well-aligned with the user's objectives and outperforms RL agents trained with reward functions learned via supervised learning. | We make reward design easier by using large language models models (like GPT-3) as a proxy for a user's reward function given that a user provides a few examples (or a description) of the desired behavior. |
We show two average-reward off-policy control algorithms, Differential Q Learning (Wan, Naik, \& Sutton 2021a) and RVI Q Learning (Abounadi Bertsekas \& Borkar 2001), converge in weakly-communicating MDPs. Weakly-communicating MDPs are the most general class of MDPs that a learning algorithm with a single stream of experience can guarantee obtaining a policy achieving optimal reward rate. The original convergence proofs of the two algorithms require that all optimal policies induce unichains, which is not necessarily true for weakly-communicating MDPs. To the best of our knowledge, our results are the first showing average-reward off-policy control algorithms converge in weakly-communicating MDPs. As a direct extension, we show that average-reward options algorithms introduced by (Wan, Naik, \& Sutton 2021b) converge if the Semi-MDP induced by options is weakly-communicating. | Showing average-reward off-policy control algorithms converge in weakly-communicating MDPs |
Fast and accurate adaptation of automatic speech recognition (ASR) systems using only text data in the target domain is a problem of long-standing practical relevance. Text-only adaptation was easy in traditional cascaded ASR systems with completely decoupled acoustic and language models. Recently, the RNNTransducer (RNN-T) has emerged as a default ASR model because of its high accuracy, low latency, and capability of supporting streaming input. However text-only adaptation of the RNN-T model is significantly more challenging due to its tight integration of acoustic and language models and end-to-end training. Existing recent approaches for text-only adaptation of RNN-Ts, either entail significant modification to the network or introduce high latency during decoding. We propose a new approach (TOLSTOI) that imputes speech representations internal to a baseline RNN-T, starting from text-only inputs, and performs in-situ adaptation that results in higher adaptation accuracy without any runtime overheads during decoding. Our imputation model is a function of the labeled data and trained parameters of the ASR model, and that we show, is more effective in controlling catastrophic forgetting compared to existing methods. We establish the effectiveness of TOLSTOI using three target domains and two ASR models of varying complexity. We yield up to 35% relative reduction in word error rate with text-only adaptation while forgetting the least compared to existing adaptation approaches. Our method is easy to implement and can be harnessed on existing RNN-T models without requiring ASR model training from scratch. | A lightweight text-only adaptation technique for end-to-end speech recognition that is both fast and accurate. |
Acquiring labeled data is challenging in many machine learning applications with limited budgets. Active learning gives a procedure to select the most informative data points and improve data efficiency by reducing the cost of labeling. The info-max learning principle maximizing mutual information such as BALD has been successful and widely adapted in various active learning applications. However, this pool-based specific objective inherently introduces a redundant selection and further requires a high computational cost for batch selection. In this paper, we design and propose a new uncertainty measure, Balanced Entropy Acquisition (BalEntAcq), which captures the information balance between the uncertainty of underlying softmax probability and the label variable. To do this, we approximate each marginal distribution by Beta distribution. Beta approximation enables us to formulate BalEntAcq as a ratio between an augmented entropy and the marginalized joint entropy. The closed-form expression of BalEntAcq facilitates parallelization by estimating two parameters in each marginal Beta distribution. BalEntAcq is a purely standalone measure without requiring any relational computations with other data points. Nevertheless, BalEntAcq captures a well-diversified selection near the decision boundary with a margin, unlike other existing uncertainty measures such as BALD, Entropy, or Mean Standard Deviation (MeanSD). Finally, we demonstrate that our balanced entropy learning principle with BalEntAcq consistently outperforms well-known linearly scalable active learning methods, including a recently proposed PowerBALD, a simple but diversified version of BALD, by showing experimental results obtained from MNIST, CIFAR-100, SVHN, and TinyImageNet datasets. | We propose a new bayesian active learning principle. |
Despite their success with unstructured data, deep neural networks are not yet a panacea for structured tabular data. In the tabular domain, their efficiency crucially relies on various forms of regularization to prevent overfitting and provide strong generalization performance. Existing regularization techniques include broad modelling decisions such as choice of architecture, loss functions, and optimization methods. In this work, we introduce Tabular Neural Gradient Orthogonalization and Specialization (TANGOS), a novel framework for regularization in the tabular setting built on latent unit attributions. The gradient attribution of an activation with respect to a given input feature suggests how the neuron attends to that feature, and is often employed to interpret the predictions of deep networks. In TANGOS, we take a different approach and incorporate neuron attributions directly into training to encourage orthogonalization and specialization of latent attributions in a fully-connected network. Our regularizer encourages neurons to focus on sparse, non-overlapping input features and results in a set of diverse and specialized latent units. In the tabular domain, we demonstrate that our approach can lead to improved out-of-sample generalization performance, outperforming other popular regularization methods. We provide insight into why our regularizer is effective and demonstrate that TANGOS can be applied jointly with existing methods to achieve even greater generalization performance. | We introduce TANGOS, a regularization method that orthogonalizes the gradient attribution of neurons to improve the generalization of deep neural networks on tabular data. |
Federated Learning is a framework for training machine learning models from multiple local data sets without access to the data in aggregate. A shared model is jointly learned through an interactive process between server and clients that combines locally learned model gradients or weights. However, the lack of data transparency naturally raises concerns about model security. Recently, several state-of-the-art backdoor attacks have been proposed, which achieve high attack success rates while simultaneously being difficult to detect, leading to compromised federated learning models. In this paper, motivated by differences in the output layer distribution between models trained with and without the presence of backdoor attacks, we propose a defense method that can prevent backdoor attacks from influencing the model while maintaining the accuracy of the original classification task. TAG leverages a small validation data set to estimate the largest change that a benign user's local training can make to the output layer of the shared model, which can be used as a cutoff for returning user models. Experimental results on multiple data sets show that TAG defends against backdoor attacks even when 40\% of the user submissions to update the shared model are malicious. | TAG is a novel defense against Backdoor Attacks in Federated Learning |
Successful and effective communication between humans and AI relies on a shared experience of the world. By training solely on written text, current language models (LMs) miss the grounded experience of humans in the real-world---their failure to relate language to the physical world causes knowledge to be misrepresented and obvious mistakes in their reasoning. We present Mind's Eye, a paradigm to ground language model reasoning in the physical world. Given a physical reasoning question, we use a computational physics engine (DeepMind's MuJoCo) to simulate the possible outcomes, and then use the simulation results as part of the input, which enables language models to perform reasoning. Experiments on 39 tasks in a physics alignment benchmark demonstrate that Mind's Eye can improve reasoning ability by a large margin (27.9% zero-shot, and 46.0% few-shot absolute accuracy improvement on average). Smaller language models armed with Mind's Eye can obtain similar performance to models that are 100x larger. Finally, we confirm the robustness of Mind's Eye through ablation studies. | We present a new reasoning paradigm that grounds language model reasoning on simulation results from the advanced physics engine MuJoCo. |
Disease is a core concept in the medical field, and the task of normalizing disease names is the basis of all disease-related tasks. However, due to the multi-axis and multi-grain nature of disease names, incorrect information is often injected and harms the performance when using general text data augmentation techniques. To address the above problem, we propose a set of data augmentation techniques that work together as an augmented training task for disease normalization, which is called Disease Data Augmentation (DDA). Our data augmentation methods are based on both the clinical disease corpus and standard disease corpus derived from ICD-10 coding. Extensive experiments are conducted to show the effectiveness of our proposed methods. The results demonstrate that our method can have up to 3\% performance gain compared to non-augmented counterparts, and they can work even better on smaller datasets. | A novel data augmentation method in NLP to address the problem of Chinese Disease Normalization. |
We show that combining human prior knowledge with end-to-end learning can improve the robustness of deep neural networks by introducing a part-based model for object classification. We believe that the richer form of annotation helps guide neural networks to learn more robust features without requiring more samples or larger models. Our model combines a part segmentation model with a tiny classifier and is trained end-to-end to simultaneously segment objects into parts and then classify the segmented object. Empirically, our part-based models achieve both higher accuracy and higher adversarial robustness than a ResNet-50 baseline on all three datasets. For instance, the clean accuracy of our part models is up to 15 percentage points higher than the baseline's, given the same level of robustness. Our experiments indicate that these models also reduce texture bias and yield better robustness against common corruptions and spurious correlations. The code is publicly available at https://github.com/chawins/adv-part-model. | Using an auxiliary task and richer annotation in the form of part segmentation can improve robustness of neural networks by a large margin. |
Language models have recently achieved strong performance across a wide range of NLP benchmarks. However, real world tasks are often poorly specified, and agents must deduce the intended behavior from a combination of context, instructions, and examples. We investigate how both humans and models behave in the face of such task ambiguity by proposing AmbiBench, a new benchmark of six ambiguously-specified classification tasks. We evaluate humans and models on AmbiBench by seeing how well they identify the intended task using 1) instructions with varying degrees of ambiguity, and 2) different numbers of labeled examples. We find that the combination of model scaling (to 175B parameters) and reinforcement learning from human feedback (RLHF) enables models to approach or exceed the accuracy of human participants across tasks, but that either one of these alone is not sufficient. In addition, we show how to dramatically improve the accuracy of language models trained without RLHF by finetuning on a small number of ambiguous in-context examples, providing a promising direction for teaching models to generalize well in the face of ambiguity. | We motivate the direction of studying task ambiguity in humans and language models, evaluating them on a new benchmark of ambiguously-specified tasks and develop methods for improving performance |
This work establishes low test error of gradient flow (GF) and stochastic gradient descent (SGD) on two-layer ReLU networks with standard initialization scale, in three regimes where key sets of weights rotate little (either naturally due to GF and SGD, or due to an artificial constraint), and making use of margins as the core analysis technique. The first regime is near initialization, specifically until the weights have moved by $\mathcal{O}(\sqrt m)$, where $m$ denotes the network width, which is in sharp contrast to the $\mathcal{O}(1)$ weight motion allowed by the Neural Tangent Kernel (NTK); here it is shown that GF and SGD only need a network width and number of samples inversely proportional to the NTK margin, and moreover that GF attains at least the NTK margin itself and in particular escapes bad KKT points of the margin objective, whereas prior work could only establish nondecreasing but arbitrarily small margins. The second regime is the Neural Collapse (NC) setting, where data lies in well-separated groups, and the sample complexity scales with the number of groups; here the contribution over prior work is an analysis of the entire GF trajectory from initialization. Lastly, if the inner layer weights are constrained to change in norm only and can not rotate, then GF with large widths achieves globally maximal margins, and its sample complexity scales with their inverse; this is in contrast to prior work, which required infinite width and a tricky dual convergence assumption.
| This work establishes low test error of gradient methods on two-layer ReLU networks with standard initialization, in three regimes where key sets of weights rotate little, making use of margins as the core analysis technique. |
Modern applications increasingly require learning and forecasting latent dynamics from high-dimensional time-series. Compared to univariate time-series forecasting, this adds a new challenge of reasoning about the latent dynamics of an unobserved abstract state. Sequential latent variable models (LVMs) present an attractive solution, although existing works either struggle with long-term forecasting or have difficulty learning across diverse dynamics. In this paper, we first present a conceptual framework of sequential LVMs to unify existing works, contrast their fundamental limitations, and identify an intuitive solution to long-term forecasting for diverse dynamics via meta-learning. We then present the first framework of few-shot forecasting for high-dimensional time-series: instead of learning a single dynamic function, we leverage data of diverse dynamics and learn to adapt latent dynamic functions to few-shot support series. This is realized via Bayesian meta-learning underpinned by: 1) a latent dynamic function conditioned on knowledge derived from few-shot support series, and 2) a meta-model that learns to extract such dynamic-specific knowledge via feed-forward embedding of support set. We compared the presented framework with a comprehensive set of baseline models trained 1) globally on the large meta-training set with diverse dynamics, and 2) individually on single dynamics, both with and without fine-tuning to k-shot support series used by the meta-models. We demonstrated that the presented framework is agnostic to the latent dynamic function of choice and, at meta-test time, is able to forecast for new dynamics given variable-shot of support series. | We present the very first step toward few-shot high-dimensional sequence forecasting by a Bayesian meta-learning model that learns the process of learning latent dynamics that changes with the small number of observations that are available. |
More data is expected to help us generalize to a task. But real datasets can contain out-of-distribution (OOD) data; this can come in the form of heterogeneity such as intra-class variability but also in the form of temporal shifts or concept drifts. We demonstrate a counter-intuitive phenomenon for such problems: generalization error of the task can be a non-monotonic function of the number of OOD samples; a small number of OOD samples can improve generalization but if the number of OOD samples is beyond a threshold, then the generalization error can deteriorate. We also show that if we know which samples are OOD, then using a weighted objective between the target and OOD samples ensures that the generalization error decreases monotonically. We demonstrate and analyze this phenomenon using linear classifiers on synthetic datasets and medium-sized neural networks on vision benchmarks such as MNIST, CIFAR-10, CINIC-10, PACS, and DomainNet, and observe the effect data augmentation, hyperparameter optimization, and pre-training have on this behavior. | We study the distribution shifts that could occur within datasets and demonstrate that under such shifts, the generalization error of the desired target task can be a non-monotonic function of the number of OOD samples. |
Most offline reinforcement learning (RL) methods suffer from the trade-off between improving the policy to surpass the behavior policy and constraining the policy to limit the deviation from the behavior policy as computing $Q$-values using out-of-distribution (OOD) actions will suffer from errors due to distributional shift. The recent proposed \textit{In-sample Learning} paradigm (i.e., IQL), which improves the policy by quantile regression using only data samples, shows great promise because it learns an optimal policy without querying the value function of any unseen actions. However, it remains unclear how this type of method handles the distributional shift in learning the value function. In this work, we make a key finding that the in-sample learning paradigm arises under the \textit{Implicit Value Regularization} (IVR) framework. This gives a deeper understanding of why the in-sample learning paradigm works, i.e., it applies implicit value regularization to the policy. Based on the IVR framework, we further propose two practical algorithms, Sparse $Q$-learning (SQL) and Exponential $Q$-learning (EQL), which adopt the same value regularization used in existing works, but in a complete in-sample manner. Compared with IQL, we find that our algorithms introduce sparsity in learning the value function, making them more robust in noisy data regimes. We also verify the effectiveness of SQL and EQL on D4RL benchmark datasets and show the benefits of in-sample learning by comparing them with CQL in small data regimes. Code is available at \url{https://github.com/ryanxhr/SQL}. | We show that some form of Implicit Value Regularization (IVR) will result in the In-sample Learning paradigm in offline RL. We also propose a practical algorithm based on the IVR framework, which obtains new SOTA results. |
We study the online restless bandit problem, where each arm evolves according to a Markov chain independently, and the reward of pulling an arm depends on both the current state of the corresponding Markov chain and the action. The agent (decision maker) does not know the transition kernels and reward functions, and cannot observe the states of arms all the time. The goal is to sequentially choose which arms to pull so as to maximize the expected cumulative rewards collected. In this paper, we propose TSEETC, a learning algorithm based on Thompson Sampling with Episodic Explore-Then-Commit. The algorithm proceeds in episodes of increasing length and each episode is divided into exploration and exploitation phases. In the exploration phase in each episode, action-reward samples are collected in a round-robin way and then used to update the posterior as a mixture of Dirichlet distributions. At the beginning of the exploitation phase, TSEETC generates a sample from the posterior distribution as true parameters. It then follows the optimal policy for the sampled model for the rest of the episode. We establish the Bayesian regret bound $\tilde {\mathcal{O}}(\sqrt{T})$ for TSEETC, where $T$ is the time horizon. This is the first bound that is close to the lower bound of restless bandits, especially in an unobserved state setting. We show through simulations that TSEETC outperforms existing algorithms in regret. | We propose TSEETC to slove the restless bandits with unknown transition kernels,unknown reward functions and unobserved states. |
Although disentangled representations are often said to be beneficial for downstream tasks, current empirical and theoretical understanding is limited. In this work, we provide evidence that disentangled representations coupled with sparse base-predictors improve generalization. In the context of multi-task learning, we prove a new identifiability result that provides conditions under which maximally sparse base-predictors yield disentangled representations. Motivated by this theoretical result, we propose a practical approach to learn disentangled representations based on a sparsity-promoting bi-level optimization problem. Finally, we explore a meta-learning version of this algorithm based on group Lasso multiclass SVM base-predictors, for which we derive a tractable dual formulation. It obtains competitive results on standard few-shot classification benchmarks, while each task is using only a fraction of the learned representations. | We show how disentangled representations combined with sparse base-predictors can improve generalization and how, in a multi-task learning setting, sparsity regularization on the task-specific predictors can induce disentanglement. |
Generating long, temporally consistent video remains an open challenge in video generation. Primarily due to computational limitations, most prior methods limit themselves to training on a small subset of frames that are then extended to generate longer videos through a sliding window fashion. Although these techniques may produce sharp videos, they have difficulty retaining long-term temporal consistency due to their limited context length. In this work, we present Temporally Consistent Video Transformer (TECO), a vector-quantized latent dynamics video prediction model that learns compressed representations to efficiently condition on long videos of hundreds of frames during both training and generation. We use a MaskGit prior for dynamics prediction which enables both sharper and faster generations compared to prior work. Our experiments show that TECO outperforms SOTA baselines in a variety of video prediction benchmarks ranging from simple mazes in DMLab, large 3D worlds in Minecraft, and complex real-world videos from Kinetics-600. In addition, to better understand the capabilities of video prediction models in modeling temporal consistency, we introduce several challenging video prediction tasks consisting of agents randomly traversing 3D scenes of varying difficulty. This presents a challenging benchmark for video prediction in partially observable environments where a model must understand what parts of the scenes to re-create versus invent depending on its past observations or generations. An anonymized website with samples can be found at https://sites.google.com/view/iclr23-teco | An efficient temporally consistent video prediction model able to generate long videos referencing hundreds of frames of past context in complex 3D environments and Kinetics-600. |
Catastrophic forgetting and the stability-plasticity dilemma are two major obstacles to continual learning. In this paper we first propose a theoretical analysis of a SPCA-based continual learning algorithm using high dimensional statistics. Second, we design OSCL (Optimized Spca-based Continual Learning) which builds on a flexible task optimization based on the theory. By optimizing a single task, catastrophic forgetting can be prevented theoretically. While optimizing multi-tasks, the trade-off between integrating knowledge from the new task and retaining previous knowledge of the old task can be achieved by assigning appropriate weights to corresponding tasks in compliance with the objectives. Experimental results confirm that the various theoretical conclusions are robust to a wide range of data distributions. Besides, several applications on synthetic and real data show that the proposed method while being computationally efficient, achieves comparable results with some state of the art. | This paper proposes a theoretical analysis of a simple but efficient continual learning algorithm |
The Transformer is an extremely powerful and prominent deep learning architecture. In this work, we challenge the commonly held belief in deep learning that going deeper is better, and show an alternative design approach that is building wider attention Transformers. We demonstrate that wide single layer Transformer models can compete with or outperform deeper ones in a variety of Natural Language Processing (NLP) tasks when both are trained from scratch. The impact of changing the model aspect ratio on Transformers is then studied systematically. This ratio balances the number of layers and the number of attention heads per layer while keeping the total number of attention heads and all other hyperparameters constant. On average, across 4 NLP tasks and 10 attention types, single layer wide models perform 0.3% better than their deep counterparts. We show an in-depth evaluation and demonstrate how wide models require a far smaller memory footprint and can run faster on commodity hardware, in addition, these wider models are also more interpretable. For example, a single layer Transformer on the IMDb byte level text classification has 3.1x faster inference latency on a CPU than its equally accurate deeper counterpart, and is half the size. Our results suggest that the critical direction for building better Transformers for NLP is their width, and that their depth is less relevant. | Widening the attention layer in a Transformer and only using a single layer is surprisingly effective, with a number of advantages. |
Real-world machine learning applications often involve deploying neural networks to domains that are not seen in the training time. Hence, we need to understand the extrapolation of \textit{nonlinear} models---under what conditions on the distributions and function class, models can be guaranteed to extrapolate to new test distributions. The question is very challenging because even two-layer neural networks cannot be guaranteed to extrapolate outside the support of the training distribution without further assumptions on the domain shift. This paper makes some initial steps towards analyzing the extrapolation of nonlinear models for structured domain shift. We primarily consider settings where the \textit{marginal} distribution of each coordinate of the data (or subset of coordinates) do not shift significantly across the training and test distributions, but the joint distribution may have a much bigger shift. We prove that the family of nonlinear models of the form $f(x)=\sum f_i(x_i)$, where $f_i$ is an \emph{arbitrary} function on the subset of features $x_i$, can extrapolate to unseen distributions, if the covariance of the features is well-conditioned. To the best of our knowledge, this is the first result that goes beyond linear models and the bounded density ratio assumption, even though the assumptions on the distribution shift and function class are stylized. | We prove the first result for understanding the extrapolation of nonlinear model class with structured domain shifts. |
Recent techniques for approximating Nash equilibria in very large games leverage neural networks to learn approximately optimal policies (strategies). One promis- ing line of research uses neural networks to approximate counterfactual regret minimization (CFR) or its modern variants. DREAM, the only current CFR-based neural method that is model free and therefore scalable to very large games, trains a neural network on an estimated regret target that can have extremely high variance due to an importance sampling term inherited from Monte Carlo CFR (MCCFR). In this paper we propose an unbiased model-free method that does not require any importance sampling. Our method, ESCHER, is principled and is guaranteed to converge to an approximate Nash equilibrium with high probability. We show that the variance of the estimated regret of ESCHER is orders of magnitude lower than DREAM and other baselines. We then show that ESCHER outperforms the prior state of the art—DREAM and neural fictitious self play (NFSP)—on a number of games and the difference becomes dramatic as game size increases. In the very large game of dark chess, ESCHER is able to beat DREAM and NFSP in a head-to-head competition over 90% of the time. | We propose a principled deep CFR algorithm that can scale to large games by removing importance sampling |
Large language models (LLMs) transfer well to new tasks out-of-the-box simply given a natural language prompt that demonstrates how to perform the task and no additional training. Prompting is a brittle process wherein small modifications to the prompt can cause large variations in the model predictions, and therefore significant effort is dedicated towards designing a painstakingly crafted "perfect prompt" for a task. To mitigate the high degree of effort, we instead ask whether collecting multiple decent, yet imperfect, prompts and aggregating them can lead to a high quality prompting strategy. Our observations motivate our proposed method, Ask Me Anything (AMA). We first develop an understanding of the effective prompt formats, finding question-answering (QA) prompts, which encourage open-ended generation ("Who went to the park?") tend to outperform those that restrict the model outputs ("John went to the park. True or False?"). AMA recursively uses the LLM to transform task inputs to the effective QA format. AM generates multiple questions per input and applies these prompts to collect several noisy "votes" for the input's true label. We find the prompts have varying accuracies and dependencies and thus propose to use weak supervision, a procedure for combining the noisy predictions, to produce the final predictions. We evaluate AMA across open-source model families (EleutherAI, BLOOM, OPT, and T0) and sizes (125M-175B parameters), demonstrating an average performance lift of 10.2\% over the few-shot baseline. This simple strategy enables the open-source GPT-J-6B model to match and exceed the performance of few-shot GPT3-175B on 15 of 20 popular benchmarks. Averaged across these tasks, the GPT-J-6B model outperforms few-shot GPT3-175B. We release our code here: https://github.com/HazyResearch/ama_prompting. | We propose a prompting strategy based on aggregating the predictions of multiple prompts, which enables a 6B parameter model to exceed the few-shot performance of GPT3-175B on 15/20 popular benchmarks. |
Due to their superior results, Transformer-based models such as BERT have become de facto standards in many Natural Language Processing (NLP) applications. However, the intensive use of complex non-linear functions within the Transformer architecture impairs its computing efficiency and complicates corresponding accelerator designs, because non-linear functions are generally computation-intensive and require special hardware support. In light of this, we propose MA-BERT, which allows matrix arithmetic-only operations in Transformer-based NLP models and achieves efficient inference with negligible accuracy loss. Specifically, we propose four correlated techniques that include approximating softmax with a two-layer neural network, replacing GELU with ReLU, fusing normalization layers with adjacent linear layers, and leveraging knowledge transfer from baseline models. Through these techniques, we are able to eliminate the major non-linear functions in Transformer-based models and obtain MA-BERT with only matrix arithmetic and trivial ReLU operations without compromising on accuracy. With mainly regular matrix arithmetic operations, MA-BERT enables hardware-friendly processing on various computing engines, including CPUs and GPUs. Our experimental results show that MA-BERT achieves up to 27% and 41% reduction in inference time on CPU and GPU, respectively, with comparable accuracy on many downstream tasks compared to the baseline BERT models. | MA-BERT completely eliminates the complex non-linear functions in BERT and achieves matrix arithmetic-only operation with trivial ReLU, which could benefit inference on both general computing units and accelerator designs for edge applications |
Frequent and structurally related subgraphs, also known as network motifs, are valuable features of many datasets. However, strong combinatorial bottlenecks have made it difficult to extract motifs and use them in learning tasks without strong constraints on the motif properties. In this work we propose a representation learning method based on learnable graph coarsening, MotiFiesta which is the first to be able to extract large and approximate motifs in a fully differentiable manner. We build benchmark datasets and evaluation metrics which test the ability our proposed and future models to capture different aspects of motif discovery where ground truth motifs are not known. Finally, explore the notion of exploiting learned motifs as an inductive bias in real-world datasets by showing competitive performance on motif-based featuresets with established real-world benchmark datasets against concurrent architectures. | An evaluation framework and model for identifying frequent subgraphs with structural flexibility in large datasets. |
Despite that going deep has proven successful in many neural architectures, the existing graph transformers are relatively shallow. In this work, we explore whether more layers are beneficial to graph transformers, and find that current graph transformers suffer from the bottleneck of improving performance by increasing depth. Our further analysis reveals the reason is that deep graph transformers are limited by the vanishing capacity of global attention, restricting the graph transformer from focusing on the critical substructure and obtaining expressive features. To this end, we propose a novel graph transformer model named DeepGraph that explicitly employs substructure tokens in the encoded representation, and applies local attention on related nodes to obtain substructure based attention encoding. Our model enhances the ability of the global attention to focus on substructures and promotes the expressiveness of the representations, addressing the limitation of self-attention as the graph transformer deepens. Experiments show that our method unblocks the depth limitation of graph transformers and results in state-of-the-art performance across various graph benchmarks with deeper models. | We analyze and solve the depth bottleneck of graph transformers from the perspective of attention capacity. |
Denoising diffusion probabilistic models are currently becoming the leading paradigm of generative modeling for many important data modalities. Being the most prevalent in the computer vision community, diffusion models have also recently gained some attention for other domains, including speech, NLP, and graph-like data. In this work, we investigate if the framework of diffusion models can be advantageous for general tabular problems, where datapoints are typically represented by vectors of heterogeneous features. The inherent heterogeneity of tabular data makes it quite challenging for accurate modeling, since the individual features can be of completely different nature, i.e., some of them can be continuous and some of them can be discrete. To address such data types, we introduce TabDDPM --- a diffusion model that can be universally applied to any tabular dataset and handles any types of features. We extensively evaluate TabDDPM on a wide set of benchmarks and demonstrate its superiority over existing GAN/VAE alternatives, which is consistent with the advantage of diffusion models in other fields. Additionally, we show that TabDDPM can be successfully used in privacy-oriented setups, where the original datapoints cannot be shared. | Proposed a new state-of-the-art approach for tabular data generation using diffusion models |
The ability to quickly and accurately identify covariate shift at test time is a critical and often overlooked component of safe machine learning systems deployed in high-risk domains. While methods exist for detecting when predictions should not be made on out-of-distribution test examples, identifying distributional level differences between training and test time can help determine when a model should be removed from the deployment setting and retrained. In this work, we define harmful covariate shift (HCS) as a change in distribution that may weaken the generalization of a predictive model. To detect HCS, we use the discordance between an ensemble of classifiers trained to agree on training data and disagree on test data. We derive a loss function for training this ensemble and show that the disagreement rate and entropy represent powerful discriminative statistics for HCS. Empirically, we demonstrate the ability of our method to detect harmful covariate shift with statistical certainty on a variety of high-dimensional datasets. Across numerous domains and modalities, we show state-of-the-art performance compared to existing methods, particularly when the number of observed test samples is small. | We propose the Detectron, a method based that detects covariate shift using an ensemble of classifiers trained to disagree with each other on unlabeled samples. |
Masked image modeling (MIM), an emerging self-supervised pre-training method, has shown impressive success across numerous downstream vision tasks with Vision transformers (ViTs). Its underlying idea is simple: a portion of the input image is randomly masked out and then reconstructed via the pre-text task. However, the working principle behind MIM is not well explained, and previous studies insist that MIM primarily works for the Transformer family but is incompatible with CNNs. In this paper, we first study interactions among patches to understand what knowledge is learned and how it is acquired via the MIM task. We observe that MIM essentially teaches the model to learn better middle-order interactions among patches and extract more generalized features. Based on this fact, we propose an Architecture-Agnostic Masked Image Modeling framework (A$^2$MIM), which is compatible with both Transformers and CNNs in a unified way. Extensive experiments on popular benchmarks show that our A$^2$MIM learns better representations without explicit design and endows the backbone model with the stronger capability to transfer to various downstream tasks for both Transformers and CNNs. | We delve deep into masked image modeling (MIM) working mechanism and propose a generic pre-training framework (A$^2$MIM) for Transformers and CNNs. |
Real-world decision-making is often associated with large and complex action representations, which can even be unsuited for the task. For instance, the items in recommender systems have generic representations that apply to each user differently, and the actuators of a household robot can be high-dimensional and noisy. Prior works in discrete and continuous action space reinforcement learning (RL) define a retrieval-selection framework to deal with problems of scale. The retrieval agent outputs in the space of action representations to retrieve a few samples for a selection critic to evaluate. But, learning such retrieval actors becomes increasingly inefficient as the complexity in the action space rises. Thus, we propose to treat the retrieval task as one of listwise RL to propose a list of action samples that enable the selection phase to maximize the environment reward. By hedging its action proposals, we show that our agent is more flexible and sample efficient than conventional approaches while learning under a complex action space. Results are also present on \url{https://sites.google.com/view/complexaction}. | Flexible reinforcement learning under complex innumerable action spaces via listwise action retrieval |
We propose a new kind of embedding for natural language text that deeply represents semantic meaning. Standard text embeddings use the vector output of a pretrained language model. In our method, we let a language model learn from the text and then literally pick its brain, taking the actual weights of the model's neurons to generate a vector. We call this representation of the text a neural embedding. With analysis of its behavior on several datasets, we confirm the ability of this representation to reflect semantics of the text. We also compare neural embeddings with GPT sentence (SGPT) embeddings. We observe that neural embeddings achieve comparable performance with a far smaller model, and that the embeddings respond to semantics differently. | We propose a new kind of embedding for natural language text that deeply represents semantic meaning. |
The flexibility and effectiveness of message passing based graph neural networks (GNNs) induced considerable advances in deep learning on graph-structured data. In such approaches, GNNs recursively update node representations based on their neighbors and they gain expressivity through the use of node and edge attribute vectors. E.g., In computational tasks such as physics and chemistry usage of edge attributes such as relative position or distance proved to be essential. In this work, we address not what kind of attributes to use, but how to condition on this information to improve model performance. We consider three types of conditioning; weak, strong, and pure, which respectively relate to concatenation-based conditioning, gating, and transformations that are causally dependent on the attributes. This categorization provides a unifying viewpoint on different classes of GNNs, from separable convolutions to various forms of message passing networks. We provide an empirical study on the effect of conditioning methods in several tasks in computational chemistry. | We unify three conditioning methods in graph neural networks in a common formulation and compare their performance on several tasks in computational chemistry. |
Prompt-based learning has emerged as a successful paradigm in natural language processing, where a single general-purpose language model can be instructed to perform any task specified by input prompts. Yet task specification in robotics comes in various forms, such as imitating one-shot demonstrations, following language instructions, and reaching visual goals. They are often considered different tasks and tackled by specialized models. This work shows that we can express a wide spectrum of robot manipulation tasks with *multimodal prompts*, interleaving textual and visual tokens. We design a transformer-based generalist robot agent, VIMA, that processes these prompts and outputs motor actions autoregressively. To train and evaluate VIMA, we develop a new simulation benchmark with thousands of procedurally-generated tabletop tasks with multimodal prompts, 600K+ expert trajectories for imitation learning, and four levels of evaluation protocol for systematic generalization. VIMA achieves strong scalability in both model capacity and data size. It outperforms prior SOTA methods in the hardest zero-shot generalization setting by up to 2.9x task success rate given the same training data. With 10x less training data, VIMA still performs 2.7x better than the top competing approach. Video demos are available at https://iclr3081.github.io/. | We design a transformer agent, VIMA, that ingests *multimodal* prompts and solves a wide variety of robot manipulation tasks. |
Deep neural networks can learn powerful prior probability models for images, as evidenced by the high-quality generations obtained with recent score-based diffusion methods. But the means by which these networks capture complex global statistical structure, apparently without suffering from the curse of dimensionality, remain a mystery. To study this, we incorporate diffusion methods into a multi-scale decomposition, reducing dimensionality by assuming a stationary local Markov model for wavelet coefficients conditioned on coarser-scale coefficients. We instantiate this model using convolutional neural networks (CNNs) with local receptive fields, which enforce both the stationarity and Markov properties. Global structures are captured using a CNN with receptive fields covering the entire (but small) low-pass image. We test this model on a dataset of face images, which are highly non-stationary and contain large-scale geometric structures.
Remarkably, denoising, super-resolution, and image synthesis results all demonstrate that these structures can be captured with significantly smaller conditioning neighborhoods than required by a Markov model implemented in the pixel domain. Our results show that score estimation for large complex images can be reduced to low-dimensional Markov conditional models across scales, alleviating the curse of dimensionality. | We develop a spatially Markov wavelet conditional probability model for images, and demonstrate (through, denoising, super-resolution and synthesis) its effectiveness in capturing global dependencies. |
Memory Gym is a novel benchmark for challenging Deep Reinforcement Learning agents to memorize events across long sequences, be robust to noise, and generalize. It consists of the partially observable 2D and discrete control environments Mortar Mayhem, Mystery Path, and Searing Spotlights. These environments are believed to be unsolvable by memory-less agents because they feature strong dependencies on memory and frequent agent-memory interactions. Empirical results based on Proximal Policy Optimization (PPO) and Gated Recurrent Unit (GRU) underline the strong memory dependency of the contributed environments. The hardness of these environments can be smoothly scaled, while different levels of difficulty (some of them unsolved yet) emerge for Mortar Mayhem and Mystery Path. Surprisingly, Searing Spotlights poses a tremendous challenge to GRU-PPO, which remains an open puzzle. Even though the
randomly moving spotlights reveal parts of the environment’s ground truth, environmental ablations hint that these pose a severe perturbation to agents that leverage recurrent model architectures as their memory.
Source Code: https://github.com/MarcoMeter/drl-memory-gym/ | Memory Gym is a novel challenge especially to memory-based agents. |
We propose a novel algorithm for offline reinforcement learning called Value Iteration with Perturbed Rewards (VIPeR), which amalgamates the pessimism principle with random perturbations of the value function. Most current offline RL algorithms explicitly construct statistical confidence regions to obtain pessimism via lower confidence bounds (LCB), which cannot easily scale to complex problems where a neural network is used to estimate the value functions. Instead, VIPeR implicitly obtains pessimism by simply perturbing the offline data multiple times with carefully-designed i.i.d. Gaussian noises to learn an ensemble of estimated state-action {value functions} and acting greedily with respect to the minimum of the ensemble. The estimated state-action values are obtained by fitting a parametric model (e.g., neural networks) to the perturbed datasets using gradient descent. As a result, VIPeR only needs $\mathcal{O}(1)$ time complexity for action selection, while LCB-based algorithms require at least $\Omega(K^2)$, where $K$ is the total number of trajectories in the offline data. We also propose a novel data-splitting technique that helps remove a factor involving the log of the covering number in our bound. We prove that VIPeR yields a provable uncertainty quantifier with overparameterized neural networks and enjoys a bound on sub-optimality of $\tilde{\mathcal{O}}( { \kappa H^{5/2} \tilde{d} }/{\sqrt{K}})$, where $\tilde{d}$ is the effective dimension, $H$ is the horizon length and $\kappa$ measures the distributional shift. We corroborate the statistical and computational efficiency of VIPeR with an empirical evaluation on a wide set of synthetic and real-world datasets. To the best of our knowledge, VIPeR is the first algorithm for offline RL that is provably efficient for general Markov decision processes (MDPs) with neural network function approximation. | A provably and computationally efficient algorithm for offline RL with deep neural networks |
We present a deep learning approach for repairing sequential circuits against formal specifications given in linear-time temporal logic (LTL). Given a defective circuit and its formal specification, we train Transformer models to output circuits that satisfy the corresponding specification. We propose a separated hierarchical Transformer for multimodal representation learning of the formal specification and the circuit. We introduce a data generation algorithm that enables generalization to more complex specifications and out-of-distribution datasets. In addition, our proposed repair mechanism significantly improves the automated synthesis of circuits from LTL specifications with Transformers. It improves the state-of-the-art by $6.8$ percentage points on held-out instances and $11.8$ percentage points on an out-of-distribution dataset from the annual reactive synthesis competition. | We present a deep learning approach for repairing sequential circuits against formal specifications given in linear-time temporal logic (LTL). |
While deep reinforcement learning (RL) agents have showcased strong results across many domains, a major concern is their inherent opaqueness and the safety of such systems in real-world use cases. To overcome these issues, we need agents that can quantify their uncertainty and detect out-of-distribution (OOD) states. Existing uncertainty estimation techniques, like Monte-Carlo Dropout or Deep Ensembles, have not seen widespread adoption in on-policy deep RL. We posit that this is due to two reasons: concepts like uncertainty and OOD states are not well defined compared to supervised learning, especially for on-policy RL methods.
Secondly, available implementations and comparative studies for uncertainty estimation methods in RL have been limited. To overcome the first gap, we propose definitions of uncertainty and OOD for Actor-Critic RL algorithms, namely, proximal policy optimization (PPO), and present possible applicable measures. In particular, we discuss the concepts of value and policy uncertainty. The second point is addressed by implementing different uncertainty estimation methods and comparing them across a number of environments. The OOD detection performance is evaluated via a custom evaluation benchmark of in-distribution (ID) and OOD states for various RL environments. We identify a trade-off between reward and OOD detection performance. To overcome this, we formulate a Pareto optimization problem in which we simultaneously optimize for reward and OOD detection performance. We show experimentally that the recently proposed method of Masksembles strikes a favourable balance among the survey methods, enabling high-quality uncertainty estimation and OOD detection while matching the performance of original RL agents. | A setup and comparison of out-of-distribution detection methods for PPO, with Masksembles as a novel, well-performing method in the setting of RL |
As machine learning (ML) algorithms are increasingly used in high-stakes applications, concerns have arisen that they may be biased against certain social groups. Although many approaches have been proposed to make ML models fair, they typically rely on the assumption that data distributions in training and deployment are identical. Unfortunately, this is commonly violated in practice and a model that is fair during training may lead to an unexpected outcome during its deployment. Although the problem of designing robust ML models under dataset shifts has been widely studied, most existing works focus only on the transfer of accuracy. In this paper, we study the transfer of both fairness and accuracy under domain generalization where the data at test time may be sampled from never-before-seen domains. We first develop theoretical bounds on the unfairness and expected loss at deployment, and then derive sufficient conditions under which fairness and accuracy can be perfectly transferred via invariant representation learning. Guided by this, we design a learning algorithm such that fair ML models learned with training data still have high fairness and accuracy when deployment environments change. Experiments on real-world data validate the proposed algorithm. | This paper presents (1) theoretical bounds for fairness and accuracy under domain generalization, and (2) a proposed model that can achieve good fairness and accuracy in an unseen target domain through invariant representation learning. |
Learning in games involves two main challenges, even in settings in which agents seek to coordinate: convergence to equilibria and selection of good equilibria. Unfortunately, solving the issue of convergence, which is the focus of state-of-the-art models, conveys little information about the quality of the equilibria that are eventually reached, often none at all. In this paper, we study a class of games in which q-replicator (QRD), a widely-studied class of no-regret learning dynamics that include gradient descent, “standard” replicator, and log-barrier dynamics as special cases, can be shown to converge pointwise to Nash equilibria. This is the starting point for our main task, which is the mathematically challenging problem of performance. In our main contribution, we quantify both conceptually and experimentally the outcome of optimal learning dynamics via average performance metrics, i.e., metrics that couple the regions of attraction with the quality of each attracting point. We provide an exhaustive comparison between gradient descent and “standard” replicator in a class of games with severe equilibrium selection problems and empirically extend our results to all dynamics in the QRD class. Our results combine tools from machine learning, game theory, and dynamical systems and provide a framework to initiate the systematic comparison of different optimal learning dynamics in arbitrary games. | Beyond convergence, average case metrics rely on regions of attraction to compare the performance of different dynamics in multi-agent games. |
Rule-based control (RBC) is widely adopted in buildings due to its stability and robustness. It resembles a behavior cloning methodology refined by human expertise. However, it is unlikely for RBC to exceed a reinforcement learning (RL) agent’s performance since it is challenging to ingest a large number of parameters during decision-making. In this paper, we explore how to incorporate rule-based control into reinforcement learning to learn a more robust policy in both online and offline settings with a unified approach. We start with state-of-the-art online and offline RL methods, TD3 and TD3+BC, then improve on them using a dynamically weighted actor loss function to selectively choose which policy should RL models learn from at each time step of training. With experiments across multiple tasks and various weather conditions in both deterministic and stochastic scenarios, we empirically demonstrate that our dynamically weighted rule-based incorporated control regularization (RUBICON) method outperforms representative baseline methods in offline settings by 40.7% in a reward settings consisting of the combination of thermal comfort and energy consumption and by 49.7% in online settings in building-RL environments. | A unified method to incorporate rule-based policy into online and offline reinforcement learning algorithm |
While machine learning models rapidly advance the state-of-the-art on various real-world tasks, out-of-domain (OOD) generalization remains a challenging problem given the vulnerability of these models to spurious correlations. We propose a balanced mini-batch sampling strategy to transform a biased data distribution into a spurious-free balanced distribution, based on the invariance of the underlying causal mechanisms for the data generation process. We argue that the Bayes optimal classifiers trained on such balanced distribution are minimax optimal across a diverse enough environment space. We also provide an identifiability guarantee of the latent variable model of the proposed data generation process, when utilizing enough train environments. Experiments are conducted on DomainBed, demonstrating empirically that our method obtains the best performance across 20 baselines reported on the benchmark. | We propose a balanced mini-batch sampling strategy to reduce spurious correlations for domain generalization. |
The impressive success of style-based GANs (StyleGANs) in high-fidelity image synthesis has motivated research to understand the semantic properties of their latent spaces. Recently, a close relationship was observed between the semantically disentangled local perturbations and the local PCA components in the learned latent space $\mathcal{W}$. However, understanding the number of disentangled perturbations remains challenging. Building upon this observation, we propose a local dimension estimation algorithm for an arbitrary intermediate layer in a pre-trained GAN model. The estimated intrinsic dimension corresponds to the number of disentangled local perturbations. In this perspective, we analyze the intermediate layers of the mapping network in StyleGANs. Our analysis clarifies the success of $\mathcal{W}$-space in StyleGAN and suggests a method for finding an alternative. Moreover, the intrinsic dimension estimation opens the possibility of unsupervised evaluation of global-basis-compatibility and disentanglement for a latent space. Our proposed metric, called Distortion, measures an inconsistency of intrinsic tangent space on the learned latent space. The metric is purely geometric and does not require any additional attribute information. Nevertheless, the metric shows a high correlation with the global-basis-compatibility and supervised disentanglement score. Our findings pave the way towards an unsupervised selection of globally disentangled latent space among the intermediate latent spaces in a GAN. | We analyze the latent space of GAN through local dimension estimation and propose a global disentanglement metric called Distortion. |
Graph Neural Networks (GNNs) have shown superior performance in Link Prediction (LP). Especially, SEAL and its successors address the LP problem by classifying the subgraphs extracted specifically for candidate links, gaining state-of-the-art results. Nevertheless, we question whether these methods can effectively learn the information equivalent to link heuristics such as Common Neighbors, Katz index, etc. (we refer to such information as heuristic information in this work). We show that link heuristics and GNNs capture different information. Link heuristics usually collect pair-specific information by counting the involved neighbors or paths between two nodes in a candidate link, while GNNs learn node-wise representations through a neighborhood aggregation algorithm in which two nodes in the candidate link do not pay special attention to each other. Our further analysis shows that SEAL-type methods only use a GNN to model the pair-specific subgraphs and also cannot effectively capture heuristic information. To verify our analysis, a straightforward way is to compare the LP performance between existing methods and a model that learns heuristic information independently of the GNN learning. To this end, we present a simple yet light framework ComHG by directly Combining the embeddings of link Heuristics and the representations produced by a GNN. Experiments on OGB LP benchmarks show that ComHG outperforms all top competitors by a large margin, empirically confirming our propositions. Our experimental study also indicates that the contributions of link heuristics and the GNN to LP are sensitive to the graph degree, where the former is powerful on sparse graphs while the latter becomes dominant on dense graphs. | We study existing state-of-the-art GNN-based link prediction methods and show that these methods can hardly learn heuristic information. Our experiments also support our analysis. |
Contrastive vision-language models (e.g. CLIP) are typically created by updating all the parameters of a vision model and language model through contrastive training. Can such models be created by a small number of parameter updates to an already-trained language model and vision model? The literature describes techniques that can create vision-language models by updating a small number of parameters in a language model, but these require already aligned visual representations and are non-contrastive, hence unusable for latency-sensitive applications such as neural search. We explore the feasibility and benefits of parameter-efficient contrastive vision-language alignment through transfer learning: creating a model such as CLIP by minimally updating an already-trained vision and language model. We find that a minimal set of parameter updates ($<$7\%) can achieve the same performance as full-model training, and updating specific components ($<$1\% of parameters) can match 75\% of full-model training. We describe a series of experiments: we show that existing knowledge is conserved more strongly in parameter-efficient training and that parameter-efficient scaling scales with model and dataset size. Where paired-image text data is scarce but strong multilingual language models exist (e.g. low resource languages), parameter-efficient training is even preferable to full-model training. Given a fixed compute budget, parameter-efficient training allows training larger models on the same hardware, achieving equivalent performance in less time. Parameter-efficient training hence constitutes an energy-efficient and effective training strategy for contrastive vision-language models that may be preferable to the full-model training paradigm for common use cases.
Code and weights at https://github.com/codezakh/LilT. | Changing a small number (<7%) of parameters in already trained language and image models can match the performance of full model training for creating CLIP-like models. |
Federated learning allows multiple clients to jointly learn an ML model while keeping their data private. While synchronous federated learning (Sync-FL) requires the devices to share local gradients synchronously to provide better guarantees, it suffers from the problem of stragglers. This is the scenario where the faster clients have to wait for the slower ones, slowing the entire training process. Conventional techniques completely drop the updates from the stragglers and lose the opportunity to learn from the data they hold, which is especially important in a non-iid setting. Asynchronous learning (Async-FL) provides a potential solution to allow the clients to function at their own pace, which typically achieves faster convergence. Since edge devices have a low compute, it is hard to train a video action recognition task on them. We present Kuiper, a variant of Async-FL, to help heterogeneous edge devices with limited resources learn a heavy model on video-action-recognition tasks with data distributed non-IID. Kuiper introduces a novel aggregation scheme, which solves the straggler problem while considering the different data distribution at different clients. Kuiper shows a 11% faster convergence compared to Oort15 [OSDI-21], up to 12% and 9% improvement in test accuracy compared to FedBuff16 [AISTAT-22] and Oort [OSDI-21] on HMDB51, and 10% and 9% on UCF101. | We develop a moderated asynchronous algorithm for training on a video action recognition task on embedded devices with mobile GPUs. |
Semi-supervised learning (SSL) provides an effective means of leveraging unlabelled data to improve a model’s performance. Even though the domain has received a considerable amount of attention in the past years, most methods present the common drawback of lacking theoretical guarantees. Our starting point is to notice that the estimate of the risk that most discriminative SSL methods minimise is biased, even asymptotically. This bias impedes the use of standard statistical learning theory and can hurt empirical performance. We propose a simple way of removing the bias. Our debiasing approach is straightforward to implement and applicable to most deep SSL methods. We provide simple theoretical guarantees on the trustworthiness of these modified methods, without having to rely on the strong assumptions on the data distribution that SSL theory usually requires. In particular, we provide generalisation error bounds for the proposed methods. We evaluate debiased versions of different existing SSL methods, such as the Pseudo-label method and Fixmatch, and show that debiasing can compete with classic deep SSL techniques in various settings by providing better calibrated models. Additionally, we provide a theoretical explanation of the intuition of the popular SSL methods. An implementation of a debiased version of Fixmatch is available at
https://github.com/HugoSchmutz/DeFixmatch | We propose a slight modification of most common semi-supervised learning methods to make them safe by debiasing their risk estimate. In particular, we apply it successfully to Fixmatch. |
Quantization approaches can minimize storage costs while decreasing the computational complexity of a model, although there is minimal study in the GNN field on quantization networks. We studied the four primary reasons why existing quantization approaches cannot be employed extensively with GNNs: (1)Quantifying the distinctions between data sources; (2)Quantifying the distinctions between data streams; (3)Quantifying the distinctions between concentrations; (4)QAT’s Limitations. Based on this, we propose GCINT, which is an efficient quantization framework prepared for GNN training. The entire forward, backward, optimizer, and loss functions are calculated using integer data. We achieved a training acceleration ratio of nearly 10× compared to FP32 Cuda Core in RTX 2080TI INT8 Tensor Core. Our quantization is independent of the dataset and weight distribution, and more than 2,000 randomized trials have been undertaken on the 8 popular GNN benchmark datasets, with all achieving errors within 1% of the FP32. | Robust and Adaptive Quantization Algorithm for Training Graph Convolution Neural Networks Using Only Integers |
We study first-order methods for constrained min-max optimization. Existing methods either require two gradient calls or two projections in each iteration, which may be costly in some applications. In this paper, we first show that a variant of the \emph{Optimistic Gradient (OG)} method, a \emph{single-call single-projection} algorithm, has $O(\frac{1}{\sqrt{T}})$ best-iterate convergence rate for inclusion problems with operators that satisfy the weak Minty variation inequality (MVI). Our second result is the first single-call single-projection algorithm -- the \emph{Accelerated Reflected Gradient (ARG)} method that achieves the \emph{optimal $O(\frac{1}{T})$} last-iterate convergence rate for inclusion problems that satisfy negative comonotonicity. Both the weak MVI and negative comonotonicity are well-studied assumptions and capture a rich set of non-convex non-concave min-max optimization problems. Finally, we show that the \emph{Reflected Gradient (RG)} method, another \emph{single-call single-projection} algorithm, has $O(\frac{1}{\sqrt{T}})$ last-iterate convergence rate for constrained convex-concave min-max optimization, answering an open problem of [Hsieh et al., 2019]. Our convergence rates hold for standard measures such as the tangent residual and the natural residual. | We propose the first single-call single-projection algorithms with optimal convergence rate for constrained min-max optimization problems in the nonconvex-nonconcave setting. |
We consider an out-of-distribution setting where trained predictive models are deployed online in new locations (inducing conditional-shift), such that these locations are also associated with differently skewed target distributions (label-shift). While approaches for online adaptation to label-shift have recently been discussed by Wu et al. (2021), the potential presence of concurrent conditional-shift has not been considered in the literature, although one might anticipate such distributional shifts in realistic deployments. In this paper, we empirically explore the effectiveness of online adaptation methods in such situations on three synthetic and two realistic datasets, comprising both classification and regression problems. We show that it is possible to improve performance in these settings by learning additional hyper-parameters to account for the presence of conditional-shift by using appropriate validation sets. | Learning hyper-parameters on an OOD validation set can improve online black-box adaptation to label-shift when there is also conditional-shift in deployment |
Recently, there has been great progress in the self-supervised pre-training of multimodal representation models that understand image and language jointly. One particularly popular application of such models is text-to-image generation, which is typically obtained via a two-stage process: in the first stage, a representation model is trained via self-supervised objectives; then in the second stage, a conditional generative decoder is trained on top of the representation to generate natural images. In this work, we aim at bringing representation learning and conditional generation together by unifying the two stages into a single model and training objective. We present UPGen, a unified pre-trained model for both representation learning and generation. UPGen is trained with a simple masked token prediction objective on a flexible mixture of image and language data. We use a pre-trained VQGAN image tokenizer to convert images into discrete tokens, then train a masked token prediction model on both paired image-text datasets and unpaired language datasets, using randomly sampled mask ratios. We show that this masked token prediction model can be directly used to generate images and language by iteratively re-masking and predicting the masked tokens. We demonstrate empirically that UPGen serves as both a good representation learning model and a generative model for both image and language. | Unified vision-language Transformer trained with masked token prediction for both representation learning and generation of image and text. |
Federated learning (FL) has made collaboration between nodes possible without explicit sharing of local data. However, it requires the participating nodes to trust the server and its model updates, the server itself being a critical node susceptible to failure and compromise. A loss of trust in the server and a demand to aggregate the model independently for oneself has led decentralized peer-to-peer learning (P2PL) to gain traction lately. In this paper, we highlight the never before exposed vulnerabilities of P2PL towards malicious attacks and how P2PL behaves differently from FL in such a malicious environment. We then present a robust defense - P2PRISM as a secure aggregation protocol for P2PL. | We highlight the vulnerabilities in peer-to-peer learning towards malicious attacks and propose a byzantine-robust defense against the same. |
Federated Learning (FL) is a promising privacy-preserving distributed learning paradigm but suffers from high communication cost when training large-scale machine learning models. Sign-based methods, such as SignSGD \citep{bernstein2018signsgd}, have been proposed as a biased gradient compression technique for reducing the communication cost. However, sign-based algorithms could diverge under heterogeneous data, which thus motivated the development of advanced techniques, such as the error-feedback method and stochastic sign-based compression, to fix this issue.
Nevertheless, these methods still suffer from slower convergence rates. Besides, none of them allows multiple local SGD updates like FedAvg \citep{mcmahan2017communication}. In this paper, we propose a novel noisy perturbation scheme with a general symmetric noise distribution for sign-based compression, which not only allows one to flexibly control the tradeoff between gradient bias and convergence performance, but also provides a unified viewpoint to existing stochastic sign-based methods. More importantly, we propose the very first sign-based FedAvg algorithm ($z$-SignFedAvg). Theoretically, we show that $z$-SignFedAvg achieves a faster convergence rate than existing sign-based methods and, under the uniformly distributed noise, can enjoy the same convergence rate as its uncompressed counterpart. Last but not the least, we remark that adding random noise to the local gradients has a double benefit: it protects the clients' privacy by, e.g., the Differential Privacy. Extensive experiments are conducted to demonstrate that the $z$-SignFedAvg can achieve competitive empirical performance on real datasets. | This work proposes the first federated averaging algorithm with sign-based compression. |
Nonconvex and nonsmooth optimization problems are important and challenging for statistics and machine learning. In this paper, we propose Projected Proximal Gradient Descent (PPGD) which solves a class of nonconvex and nonsmooth optimization problems, where the nonconvexity and nonsmoothness come from a nonsmooth regularization term which is nonconvex but piecewise convex. In contrast with existing convergence analysis of accelerated PGD methods for nonconvex and nonsmooth problems based on the Kurdyka-\L{}ojasiewicz (K\L{}) property, we provide a new theoretical analysis showing local fast convergence of PPGD. It is proved that PPGD achieves a fast convergence rate of $O(1/k^2)$ when the iteration number $k \ge k_0$ for a finite $k_0$ on a class of nonconvex and nonsmooth problems under mild assumptions, which is locally the Nesterov's optimal convergence rate of first-order methods on smooth and convex objective function with Lipschitz continuous gradient. Experimental results demonstrate the effectiveness of PPGD. | We propose Projected Proximal Gradient Descent (PPGD) which solves a class of non-convex and non-smooth optimization problems with the Nesterov's optimal convergence rate. |
Real-world data usually couples the label ambiguity and heavy imbalance, challenging the algorithmic robustness of partial label learning (PLL) and long-tailed learning (LT). The straightforward combination of LT and PLL, i.e., LT-PLL, suffers from a fundamental dilemma: LT methods build upon a given class distribution that is unavailable in PLL, and the performance of PLL is severely influenced in long-tailed context. We show that even with the auxiliary of an oracle class prior, the state-of-the-art methods underperform due to an adverse fact that the constant rebalancing in LT is harsh to the label disambiguation in PLL. To overcome this challenge, we thus propose a dynamic rebalancing method, termed as RECORDS, without assuming any prior knowledge about the class distribution. Based on a parametric decomposition of the biased output, our method constructs a dynamic adjustment that is benign to the label disambiguation process and theoretically converges to the oracle class prior. Extensive experiments on three benchmark datasets demonstrate the significant gain of RECORDS compared with a range of baselines. The code is publicly available. | We propose a novel method RECORDS for long-tailed partial label learning, which overcomes the drawback of the straightforward combination between long-tailed learning and partial label learning, and significantly improves the performance. |
The pursuit of long-term fairness involves the interplay between decision-making and the underlying data generating process. In this paper, through causal modeling with a directed acyclic graph (DAG) on the decision-distribution interplay, we investigate the possibility of achieving long-term fairness from a dynamic perspective. We propose Tier Balancing, a technically more challenging but more natural notion to achieve in the context of long-term, dynamic fairness analysis. Different from previous fairness notions that are defined purely on observed variables, our notion goes one step further, capturing behind-the-scenes situation changes on the unobserved latent causal factors that directly carry out the influence from the current decision to the future data distribution. Under the specified dynamics, we prove that in general one cannot achieve the long-term fairness goal only through one-step interventions. Furthermore, in the effort of approaching long-term fairness, we consider the mission of "getting closer to" the long-term fairness goal and present possibility and impossibility results accordingly. | We formulate and investigate a long-term fairness notion that captures decision-distribution interplay via a detailed modeling over both observed and latent causal factors. |
Recent works on neural contextual bandits have achieved compelling performances due to their ability to leverage the strong representation power of neural networks (NNs) for reward prediction. Many applications of contextual bandits involve multiple agents who collaborate without sharing raw observations, thus giving rise to the setting of federated contextual bandits}. Existing works on federated contextual bandits rely on linear or kernelized bandits, which may fall short when modeling complex real-world reward functions. So, this paper introduces the federated neural-upper confidence bound (FN-UCB) algorithm. To better exploit the federated setting, FN-UCB adopts a weighted combination of two UCBs: $\text{UCB}^{a}$ allows every agent to additionally use the observations from the other agents to accelerate exploration (without sharing raw observations), while $\text{UCB}^{b}$ uses an NN with aggregated parameters for reward prediction in a similar way to federated averaging for supervised learning. Notably, the weight between the two UCBs required by our theoretical analysis is amenable to an interesting interpretation, which emphasizes $\text{UCB}^{a}$ initially for accelerated exploration and relies more on $\text{UCB}^{b}$ later after enough observations have been collected to train the NNs for accurate reward prediction (i.e., reliable exploitation). We prove sub-linear upper bounds on both the cumulative regret and the number of communication rounds of FN-UCB, and empirically demonstrate its competitive performance. | We introduce federated neural-UCB, which uses a weighted combination of two UCBs that respectively, (a) accelerates exploration using observations from other agents and (b) improves reward prediction using a neural network with aggregated parameters. |
Knowledge distillation (KD) is a popular model compression technique to transfer knowledge from large teacher models to a small student model. Typically, the student learns to imitate the teacher by minimizing the KL divergence of its output distribution with the teacher's output distribution. We argue that such a learning objective is sub-optimal because there exists a discrepancy between the teacher's output distribution and the ground truth label distribution, and forcing the student to blindly imitate the unreliable teacher output distribution leads to inferior performance. To this end, we propose a novel knowledge distillation objective PTLoss by first representing the vanilla KL-based distillation loss function via a Maclaurin series and then perturbing the leading-order terms in this series. This perturbed loss improves the student generalizability by effectively distilling knowledge from a shifted distribution closer to the ground truth data. We also propose a method to compute this shifted teacher distribution, named Proxy Teacher, which enables us to select the perturbation coefficients in PTLoss. We theoretically show the perturbed loss reduces the deviation from the true population risk compared to the vanilla KL-based distillation loss functions. Experiments on three tasks with teachers of different scales show that our method significantly outperforms vanilla distillation loss functions and other perturbation methods. | We propose a perturbed loss function for the knowledge distillation task which outperforms the underlying KL loss and other perturbation methods. |
This paper presents a new paradigm to cluster incomplete vectors using subspaces as proxies to exploit the geometry of the Grassmannian. We leverage this new perspective to develop an algorithm to cluster and complete data in a union of subspaces via a fusion penalty formulation. Our approach does not require prior knowledge of the number of subspaces, is naturally suited to handle noise, and only requires an upper bound on the subspaces’ dimensions. In developing our model, we present local convergence guarantees. We describe clustering, completion, model selection, and sketching techniques that can be used in practice, and complement our analysis with synthetic and real-data experiments. | We introduce a novel paradigm that optimizes on the Grassmannian to complete and cluster incomplete data in a union of subspaces. |