About proximal dendritic segments


I’d like to know how many proximal dendritic segments a column has.
Is it just one the connects to a single part of the input area?
I can’t find a good explanation about it with some pictures.


In HTM, a minicolumn is a group of cells which share a proximal dendrite. The idea is, in real life, each cell in a minicolumn has pretty much a duplicate proximal dendrite, with the same inputs.

A minicolumn’s proximal dendrite receives from any input to the sequence memory (from the encoder). But if you’re talking about the thousand brains theory, there’s also cortical columns, which are different from minicolumns. Each cortical column contains hundreds of minicolumns. A cortical column receives from a patch on the sensor, e.g. skin. Each minicolumn’s proximal dendrite receives different inputs, but they’re all from within that patch.


Thanks that help clear some things up for me. But to be clear most of the talks and especially the HTM school youtube videos are only talking about minicolumns at L4 right (and your answer seems to focus mostly on L4) and other layers may have minicolumns with different functions compared to L4? And if mentions of temporal pooling/memory is involved, minicolumns at L2/L3 is vaguely implied but I don’t think how the different minicolumns at all layers of a cortical column function together have been described by HTM? I haven’t read TBT yet so maybe I’m wrong.

I think they’re not really about specific layers. For example, lower L3 might be for sequence memory, because it receives sensory inputs directly from thalamus like L4 (also some parts of other layers do too). You’re right that other layers could use minicolumns for other things. As far as roles of specific layers in HTM, it’s mostly tentative, so it’s not at the point where they’re certain about what minicolumns in each layer do.

1 Like
1 Like