Can TM or extension of TM be used to replay sequences given TP-SDR?

In general TM predicts the next step.

Is there a way using TM alone or with combination with extensions to force it to replay “stored/pooled” sequences ?
(Yes I mean temporal unfolding)

How does the brain trigger execution of sense-motor sequences ?

m’I oversimplifying to expect that a TPooled-SDR to be reversible … or it can just be used to compare quickly one sequence with another ?

It seem to me that chained TM-TP will build bottom-up hierarchy , but what about the reverse direction OR is it second separate path that is still TM-TP but in reverse direction ?

Yes, that is exactly what the TM does. That is its primary function.

Either from sensory input, internal state changes, or active thinking (playing out mental movements through simulated space).

Sorry, I don’t understand your other questions.

I assume you are talking about temporal unfolding. Assuming you find a good TP algorithm that forms stable representations for sequences (or parts of sequences) learned by TM, then one way you could do temporal unfolding is to form apical feedback connections from the TP layer back to the TM layer (using exactly the same algorithm as learning distal connections in TM).

Then you could set up a system where any cell which receives both apical and distal input will become active. Activating the first couple of elements of the sequence would then trigger the rest of the sequence to unfold. I don’t know about the biological plausibility of this strategy though.

3 Likes

There was a discussion about temporal unfolding some time back on this thread, which you might find relevant.

1 Like

This sounds a lot like what 's happening here if I understand it:

1 Like