Episode 26: Sugandha Sharma, MIT, on biologically inspired neural architectures, how memories can be implemented, and control theory
Generally Intelligent
00:00
The Learning Mechanism of the Attributor Manifold
There was a new network that actually gave a memory continuum. There's no constraint on the features. It can be any factor. Or smoke and strain. And so to sign up, she had this two-layer network that formed the underlying attractor manifold. And now you're adding this third layer. The way to think about this layer is it's kind of storing relationships between whatever features you're trying to learn and this, these predefined fixed points.
Transcript
Play full episode
