Similarity Alignment: A Missing Link Between Structure, Function and Algorithms

Feedback alignment in the human brain:
Deep Learning with Ensembles of Neocortical Microcircuits.

More nails for the DL hammer. In any case, thanks for the link.

@rhyolight sorry for lifting the old topic. I have recently followed other works by Mitya Chklovskii. He is doing really interesting research, especially on the fruit fly brains. Is it somehow possible to make an interview with him?
I know it’s not so easy. Just would like to point to this possibility. Thanks.

1 Like

I don’t have plans of interviewing neuroscientists unless they have something to do with that thalamus or cortex or some higher level brain parts than the fly brain.

1 Like

You may want to read his papers - he has some very interesting views on local voting to build representations.

This is work on an extremely important problem - how to separate different objects in a local neighborhood. This is close to the Numenta work on objects in columns. Think “cocktail party effect” to get some idea what to expect:
Blind nonnegative source separation using biological neural networks

In this he raises objections to the base neural model used in deep learning:

And suggests a better alternative based on bio-similarity here:

And a dive into principle component analysis:

“A key step in insect olfaction is the transformation of a dense representation of odors in a small population of neurons - projection neurons (PNs) of the antennal lobe - into a sparse representation in a much larger population of neurons” Very relevant to encoder design.

What do K-means clustering and manifold learning have in common? A semi-definite program formulation:

This work is relevant to the current efforts to build object representations at the column level.

1 Like