title
stringlengths
12
112
published
stringlengths
19
23
url
stringlengths
28
28
video_id
stringlengths
11
11
channel_id
stringclasses
5 values
id
stringlengths
16
31
text
stringlengths
0
596
start
float64
0
37.8k
end
float64
2.18
37.8k
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3680.28
I'm gonna just briefly talk about this. And he trashes the neuro symbolic people a bit
3,680.28
3,694.92
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3686.44
like he trashes the people that say no, no, you know, neural networks can never do whatever.
3,686.44
3,700.68
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3694.92
And he says pretty clearly look, neural networks can represent trees, I've given you a system
3,694.92
3,708.72
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3700.6800000000003
also BERT can output parse trees. So shut up, I guess. And he comes up with this glom
3,700.68
3,716.04
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3708.72
BERT name, which, you know, is is already coined. If you wanted to do glom BERT, that's
3,708.72
3,728.92
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3716.04
already taken Sorry. I also by the way, also coined the I coined the name may glow mania
3,716.04
3,733.96
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3728.9199999999996
right now. Okay, if you want to, if you want to use it, it better be a pretty cool machine
3,728.92
3,742.44
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3733.96
learning system and be based on glom. Right? That was the paper. I think it's a cool system.
3,733.96
3,747.28
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3742.44
It has a bunch of parts that are maybe not super friendly to hardware at the time like
3,742.44
3,752.24
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3747.28
this iterative procedure. But honestly, it is not much more than a neural network. Sorry,
3,747.28
3,759.36
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3752.2400000000002
a recurrent neural network with very complicated recurrence functions. The video extension
3,752.24
3,765
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3759.36
might be a bit tricky. And, but the rest and the regularization might be a bit tricky,
3,759.36
3,769.8
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3765.0
the exact objective. So the denoising auto encoder objective isn't super detailed in
3,765
3,776.36
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3769.8
the paper, he simply says, reconstruct a corrupted version of the input. How exactly the input
3,769.8
3,782
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3776.36
happens, maybe there's a CNN, maybe the CNN feeds information into actually multiple layers.
3,776.36
3,788.92
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3782.0
None of that is exactly specified. So there's lots to figure out. I do think the ideas are
3,782
3,797.32
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3788.92
very cool. And I love idea papers. And therefore I recommend that if you're interested more,
3,788.92
3,802.6
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3797.32
give this thing a read, give this video a like, share it out. And I'll see you next
3,797.32
3,819.72
GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)
2021-02-27 15:47:03
https://youtu.be/cllFzkvrYmE
cllFzkvrYmE
UCZHmQk67mSJgfCCTn7xBfew
cllFzkvrYmE-t3802.6
3,802.6
3,819.72