HTM expressed and sets and topology!

Deep learning is all about statistics and high-dimensional linear algebra.

This really does not fit the HTM model.

I have an intuition that a better tool would be some branch of set theory, with each synapse being a connection between the set of input vectors and output vectors. I suspect that it may be necessary to use this to map the high-dimension manifolds formed as maps are connected.

A necessary step is the evolution of these sets over time.

1 Like