Enhancement: TM accept external predictive input

[ctrl-z-9000-times]

The temporal memory should accept predictive inputs (onto distal basal dendrites) from an external source.

There is a python example of this at: https://github.com/numenta/htmresearch/blob/master/htmresearch/algorithms/apical_tiebreak_temporal_memory.py

[breznak]

@ctrl-z-9000-times do you want to work on this feature, can I assign it to you?

It would be best if the implementation could be just an optional addition to the current TM impl, which I believe should be possible?

[ctrl-z-9000-times]

That sounds good @breznak, though I probably won't start work on this until I've wrapped up SP-topology enhancements (#86), so it may be a while until I get around to this...

[ctrl-z-9000-times]

I would use this feature in the following ways:

• vertical connections between cortical layers. Cortical layers 2/3 project to layer 4's distal dendrites.
• Horizontal connections between distant areas of the cortex.
• Efferent motor commands use this.

[breznak]

😍 I'm very much into this.

vertical connections between cortical layers. Cortical layers 2/3 project to layer 4's distal dendrites

I'd like to re-introduce the "topDownCompute" feature, which is a feed-back connection to HTM. It would allow us to get rid of *Classifier (which is imho an artificial concept), and offer more informed & natural way of re-constructing things (SDRs, inputs) ...I'll create an Issue for this.

Horizontal connections between distant areas of the cortex.

Interesting, so something like hubs & connectomes? Would #156 use this implementation?

[ctrl-z-9000-times]

This feature allows the TM to be used as a tool for experimenting with minicolumns.

I dont envision #156 using this, since i have some rather different ideas about Apical dendrites.

[ctrl-z-9000-times]

Implemented in PR #180