Episode 26: Sugandha Sharma, MIT, on biologically inspired neural architectures, how memories can be implemented, and control theory
Generally Intelligent
00:00
Student-Verse Learning - Continuous Degradation
We're using student-verse learning for the hetero association. Depending on how big your hidden layer is, you can reconstruct some up to some features perfectly. And beyond that, your features are going to degrade in the amount of information they have. So if you were trying to store less number of patterns, then it's equivalent to saying that you're now projecting a lower dimensional pattern onto also a lower dimensional plane. But once you have large number of patterns,. then you start losing some information.
Transcript
Play full episode
