For your information an delight a lightweight summary of Neuron article:
37 synapse subcategories identified, typical synapse diversity in cortex.
2 Likes
Reading the original paper: https://www.cell.com/neuron/pdfExtended/S0896-6273(18)30581-6
What does the image on page 15 mean in the context of HTM, is it suggesting multiple dimensions to an SDR?
3 Likes
I have been mentioning this in regards to emotions for a long time now.
If you take sub-cortical generated chemical messengers as part of a system you can combine them with these synapses to activate collections of neural assemblies in response to system wide signals from the limbic system.
From the receiving end - an action potential is an action potential - all should work the same way to signal something. What changes is that some signalling is gated by what chemical co-factors are circulating in the blood and cerebral-spinal fluid.
From a system point of view what you get is essentially a sensor that is activated by internal chemicals rather than some externally sensed pattern. It does not matter that this was a chemically gated signal or a photon-generated signal - the pattern is learned either way.
One note to your implied question - most regions/areas/maps in the brain are penetrated by signalling axons from more than one other region/area/map in the brain. What the SDRs are learning is coincidences of these signals. These may be different features of some sense such as “color AND stereo disparity” or between hierarchy such as “low level raw pixel AND higher level derived curve shape”
It makes perfect sense that some shapes or sequences will be colored with a GOOD/BAD chemical signalling as part of what is learned.
It also makes sense to me that the internal body signals such as hunger or thirst modify what is sensed and processed as a form of gating of attention.
None of this chemical gating is really incorporated in the current HTM models beyond what you chose to do with your encoder.
2 Likes
This is a very important paper. I’m extremely impressed. At this point I can say that it’s describing what I have been searching for in regards to what the CA1 data looks like when displayed. This is something we need to understand.
There is so much new information and (synaptic) models to study that it might take me a month to adsorb. But I’m (in my free time) working on it!
2 Likes
Fascinating videos I found that go with the paper. The entire brain changes state?
This one takes time to load and may skip, but it’s worth a try:
1 Like