Any theories for minicolumns across all layers?


I’ve been pondering why a mini-column might span all layers in a column as mentioned in one of the 2019 research chats with Jeff and Matt, I have one possible theory, below, and am interested if people have been taking this idea and wildly theorizing about it.

HTM requires mini-columns for learning temporal patterns.

Maybe they are also important for playback of learned motor sequences? I don’t know if a single column can learn motor sequences and play them back? Is it possible that a single column can learn how to catch a ball?
If so, could it be that the mini-column is priming cells in the layers that deal with sending motor signals to continue the action of catching a ball as it has been learned?

1 Like