Object / concept representation in high-dimensional space


I think about this a lot; pretty much every time I consider the implications of tracts and maps in the connectome.

Do you have any pointers to research on this?

Importance of a Modular HTM Layer

No, this is really hard to get in your head. You have to think about it long and hard enough to develop an intuition about it. Understanding high-dimensional systems and attractors within dynamical systems helped me hone my understanding. I sometimes think of an idea as an high-dimensional attractor in the brain that can be shaped and mutated to merge and combine with other ideas and concepts.


You have been more helpful than you know; sometimes knowing what question to ask is the most important thing.
Google "high-dimensional systems and attractors within dynamical systems"
There is some good stuff there.

I am really liking this one on the first read:


Great! In my next “Interview with a Neuroscientist”, I’ll talk about dynamical systems and attractors with Dr. Jonathan Michaels from Stanford’s Neural Prosthetic Systems Lab.


I can see the immediate application in the oscillatory lower brain stem systems I have been studying.

It would be interesting to see if there is also a tie to “static” HTM/SDR model.

The thalamocortical oscillation and inter-map recruitment come to mind in the cortex but I don’t see how the HTM/SDR system could incorporate this insight.

Deep Predictive Learning: A Comprehensive Model of Three Visual Streams

Are you talking about “static” vs oscillating? Because we’re researching oscillations for inference / learning phase. See latest hangout at HTM Hackers’ Hangout - Dec 1, 2017.


I see the main-line numenta/Nupic model as a peristaltic system with a cycle of predict/perceive/post-results in each map.

I see much of the lower brain stem as a system of coupled oscillators.


This discussion reminds me of the Feynman Machine model. The authors make some claims about support in Taken’s theorem on manifold projections in dynamic systems.

And this guy seems to be on a similar track: http://blog.piekniewski.info/2016/11/01/statistics-and-dynamics/