Episode 26: Sugandha Sharma, MIT, on biologically inspired neural architectures, how memories can be implemented, and control theory
Generally Intelligent
00:00
How to Represent Non-Parametric Distributions in Neural Circuits
In order to do Asian inference, you need to represent probability distributions in neural circuits. No one had thought about representing non-parametric distribution on the neural level given a limited number of resources. The solution I have proposed was basically use a singular value decomposition to project this high dimensional space and then perform computations in this low dimensional space. So it was a simple projection of a high dimensional space into a lower dimensional space.
Transcript
Play full episode
