What are nodes in the HTM theory?

I’ve seen the usage of the term “node” in a few (usually old) writings on the HTM theory (e.g. you can find the term in the D. George’s Ph.D. thesis). What are they? Do nodes still exist in the current HTM theory? If not, what were they used for or what did they represent, and why were they “removed” from the theory? It seems to me that they were the smallest computational units of the old theory, so they would somehow be compared to the current cortical columns?!


The only way I think about “nodes” are within a hierarchy. When you think about a hierarchy of regions of cortex, you would say that a region is a node in the hierarchy. Each region would contain many cortical columns. All regions or nodes receive feedforward input. I actually drew a picture of this hold on…

This was just a brainstorming drawing I made when working on some visual concepts to illustrate the idea of the “Thousand Brains” model.

Anyway, it shows 7 nodes.


My drawing was not to imply any specific number of nodes, or how they are connected. It was only meant to illustrate that a node contains a region of cortex, which contains many cortical columns. How they are wired together between each other is all up in the air.


I had read (somewhere) that a level of a hierarchy corresponds to a region, but this is not (necessarily or always) the case. Apparently, a level of the hierarchy can have multiple regions (or, as you say, nodes). I was confused by this misleading info (I had read).

Do you mean a node corresponds to a region of the cortex?

Sure. But this is all very ambiguous in the brain. It is hard to define where borders are even if we know what they connect to. The hierarchy is much messier than you think.