This is a discussion channel for “Mixed Modular Codes and Remapping for Highly Generalizable Learning and Inference” by Prof. Ila Fiete (Massachusetts Institute of Technology). A link to the talk is the following. Please do not share the link to anyone outside of this slack workspace. Access Passcode can be found at the announcement channel.   URL: https://vimeo.com/471273721 （35 minutes）

Question to Ila-san. Thanks for great talk. I will be a chair of this session. You mentioned an importance of mixed modular codes for highly generalizable learning and inference and showed some examples of its efficacy. Do you think some specific network structure and/or constraint is required to realize such mixed modular coding, rather than fixed factorized coding?  If yes, what kind of structure is essential to realize such function ? and which brain area or which neural circuit mechanism is most comparable or suitable to represent the mixed modular coding?

An observation: phase code representation of distance in space is the most natural representation of distance. Euclid cursed us with perception of absolute space, but as Einstein has shown, it is not absolute nor straight. Einstein starts his argument in “Electrodynamics of Moving Bodies” (the paper that introduced Special Relativity) by saying that you need a protractor and ruler in order to measure space. That’s the only way you can really know how much space is there. In case of an animal the protractor and ruler is the body itself and place cells or other location-sensitive neurons are ticks on the protractor/ruler. So as we measure space with our bodies, these periodic representations are what allow us get an accurate representation of space in the brain. Sadly in our minds we live with a conception that space is Euclidean. We perceive space as Euclidean because we believe it is such—we have an incorrect model. Euclid’s Elements, the foundational text of mathematics, is also a kind of a curse. 100 years after Einstein we still have not internalized the fact that space is curved and we think that Einstein’s discovery of the more accurate geometry of space (and time) does not apply to our perception.

*Thread Reply:* To what extent is space non-Euclidean, and physics non-Newtonian, at the spatiotemporal scales of animal behaviour?

*Thread Reply:* I think the question should be asked the other way around: space simply is non-Euclidean. To what extent Euclidean space is a good approximation of space? What is space in the first place? What does it mean to occupy a position in space, what's the extent of the body in space, what does it mean to move the body from one position of space to another? In the Western intellectual tradition we think of space as emptiness, a container for things. In the Eastern contemplative traditions space is the fundamental substrate out of which other elements that make up the universe emerge.

*Thread Reply:* The textbook physics answer is that you can ignore curvature of space due to mass other than the planetary mass in calculations and consider space locally flat

*Thread Reply:* But then again, nothing around us is actually flat... https://www.youtube.com/watch?v=cr9Z9rWjoV4

YouTube

} Nguyen Hao (https://www.youtube.com/channel/UCCpNOHYKiiHsboqjwCJEI0A)

Experience True Level (Rick and Morty Season 3)

Original URL: https://www.youtube.com/watch?v=cr9Z9rWjoV4

Dear Ila,  you mentioned manifolds such as the 2D torus. This indicates that decoding and identification of higher-dimensional manifolds with non-trivial topology may become important. What type of topologies and what dimensionality do you envision?  Kind regards, Stephan

Do amnesics without a hippocampus have problems with motor control?

Interesting Q, Terry: the question is whether new motor learning is impeded. There is evidence from eyeblink conditioning that once the delay between a conditioned stimulus and unconditioned stimulus grows, that PFC and hippocampus become important for generating the response beyond cerebellum. Here we could ask about whether the role of PFC and hippocampus go beyond providing a simple memory trace.