Does HTM have an explanation for chunking, magnification, and remapping in the cortex? If so, how do these behaviors come about?

[jsandeep316]

There are three phenomena in the brain that fascinate me: chunking, magnification, and cortical remapping. I’m not sure if all of them can be ascribed to activity within the neocortex, but from my limited understanding, I think this is so.

How does HTM explain them?

[Andrew_Stephan]

I’m still new myself, but I believe chunking can be explained within HTM. Thoughts and inputs are represented as SDRs moving up and down the hierarchy. If a given collection of concepts are presented simultaneously often enough, a subset of minicolumns (neurons) would become sensitive to that collection, and thereafter those neurons would represent the new chunk.

[Bitking]

I can’t answer most of your question but I will tackle the chunking part. It is an emergent property of system level operation.

The current “contents of consciousness” - the collection of interlocking maps in the brain, each holding some part of the parsed here and now; a basket of interlocking features that combine to describe some unique collection that we take as a chunk. You have two halves of your brain that mostly are duplicates of each other and it does not take the entire brain to form a content. Humans seem to be stuck at about 4 or so simultaneous patterns before they interfere with each other. I think that this is related to how we process tuples in the loop of consciousness.

I take is as a matter of faith that the brain processes tuples serially.

(Tuple = object-relationship-object)

Thing one is recalled/perceived and the loop of consciousness projects thing two with some relationship object to be perceived, evolving to thing three.

See “loop of consciousness”:

Intelligence vs Consciousness

My working theory on consciousness is fairly simple considering how complex the central nervous system is. Let me set the table on how I see the brain working: Focusing on the H of HTM. (Hierarchy) Consider this in a sensory stream general framework of (raw sense) to (collected sense/association) to (name/symbol) to (collected symbols/grammar) to (working memory) to (the emotionally colored here and now) in the tip of the temporal lobe. This is the stream of sensation that you know as your pe…

See “contents of consciousness”:

HTM Working Memory

@Sparse HTM memory works by pattern completion. Many neural models work the same way and there is a considerable amount of theory to back up the utility of this computing model. Very few (none?) also handle temporal memory Each map in the sheet of columns has (potentially) a vast number of overlapping learned temporal/spatial patterns. When the perception is registered the saved pattern that most closely resembles the sensed pattern is recalled. There is some parsing to distribute these memori…

and the relation of the “contents of consciousness” and long term memory:

HTM Working Memory

That is a tricky point to answer. The bulk of the cortex learns by “simple” Hebbian process - relatively slow. The fact that patients like HM are capable of learning skills like mirror writing without any memory of having done so reflects that the episodic memory is what is missing when the EC/HC is compromised. Due to the distributed and interlocking nature of the representation, the cortex capacity is enormous. To understand this; there are only a small number of letters in an alphabet but d…

[SeanOConnor]

What is the logic of learning exceptions? I ask myself.
1/ Exceptions are rarer than the more general thing. By chance you will generally learn them later. You capture the conditional probabilities in a crude way.
2/ There is a process of inclusion and exclusion going on. You learn an exception but it is likely too broad and you must learn what to exclude from that exception.
3/ You get a list of alternative answers by following the chain of exceptions. You don’t have to fully accept what was seen in the data at the most extreme exception.
4/ You may choose only to add an exception only after it has been encountered a number of times. Like synapse building only after a number of repeat events.
5/ You can extract sequences of exceptions as trigger patterns, that maybe you can do further learning from. As in going from letter representations to word representations.