Grids into Maps!

[Gary_Gaulin]

@Bitking

You see that this is where the fact’s don’t support the concept. The videos that I have seen show the waves sweeping over the cortex without any diversions.

That can be true for some areas. What I see in the network are often spirals and complex waves similar to these:

‘Blue’ voltage-sensitive dyes for studying spatiotemporal dynamics in the...

Totally stopping the a traveling wave would require a large number of places leaving no holes to get through, otherwise the wave goes around all obstacles. Also from what I read a barrier/boundary cell becomes active when near one, is not always active.

@Bitking

Thinking about this a little more deeply - the contents of individual maps/areas is a distributed representation. It is fair to say that the representation is really distributed up and down the entire hierarchy but I will restrict that to a single map/area for this post. There is not a local “wall” to bounce off of. The wave has to sweep across the entire map/area to do whatever processing that happens in that area of cortex.

This would be what the navigational part of the distributed system sums up to, where there is one or many blobby representations of movement, instead of detailed picture somewhere on the cortex. To record one one the animal would have to at least in its mind be navigating a complex environment with tunnels or other features that stand out in the wave pattern.

[mthiboust]

Concerning the excitatory cells, I was referring to L5 PT and L6 CT cortical pyramidal cells that project to the thalamus, and to the miscalled thalamic “relay” cells (“miscalled” because they probably do more than a simple relay)

It seems that virtually all axons between the thalamus and the cortex give some collaterals in the TRN (which sends inhibitory inputs to thalamic relay cells and/or other inhibitory thalamic interneurons).

This illustration from Sherman’s video shows the different options:

Other illustrations from the same video:

I hope it helps. On my side, I don’t know much about the internal thalamic computations.

[Bitking]

I see I am doing a poor job of communicating the contents of maps/areas. You are describing things as objects in some map/area as if they were present as discrete objects distributed over the area like an overhead picture of the scene in a photograph. It just does not work that way at all.

The eye is an approachable path as we can form some mental models of how what we see is represented in cortex. The other senses work the same way but we will stick with the eye here. Even a quick overview shows that the idea of a wall in the cortex simply does not make sense. What we “look at” to form the concept of a wall is a sequence of fixations on the features of the world. I offer this bit about “just” looking at a face:

Saccades key to vision?

I hate to break it to you but Saccades don’t work that way. Check this out to get some idea of how real vision takes in the visual world. Here is what you “see” when you look at this picture: [Biker] Here is what the V1 area is fed with saccades driving foveal vision: [Biker%20left%20eye%20area%20-%20eye%20straight%20on%20-%20Copy]

Note that the cortex is getting the images from each fixation as a sequences, each stacked one on the next. The fixations that is part of a wall or door a occupy the exact same space in cortex. What would a wave bounce off of?

It’s harder than that as the dynamics of a traveling wave are not synchronized to what parts the eye is pointing at - the door or the table in the way or … whatever.

The assemblage of features from different fixations are assembled into objects in the temporal lobe but these cortical waves that drive processing happen over the entire cortex - not just the bits where features turn into object and relationships between objects.

[mthiboust]

@Gary_Gaulin: I feel lost when trying to understand your experiments, even after reading your discussion with @Bitking

Could you state again what you would like to demonstrate concerning the travelling waves? What parts are facts, what are speculations, and how your experiments support (or will support) your speculations?

Can we reliably associate travelling waves (that we can measure with EEG) with L2/3 activity? I guess so, but no so sure.

[Gary_Gaulin]

@mthiboust

Could you state again what you would like to demonstrate concerning the travelling waves? What parts are facts, what are speculations, and how your experiments support (or will support) your speculations?

The opening post of this thread to explain what I can has the paper “Dynamic Grouping of Hippocampal Neural Activity During Cognitive Control of Two Spatial Frames” I modeled from:

Oscillatory “Thousand Brains” Mind's Eye For HTM?

Even though I have evidence that the oscillatory dynamics of at least the hippocampal end of the sheet are eventually vital to include in computations it works so well with HTM theory I’m hopeful you’ll none the less take all this as good news. The “Thousand Brains Model of Intelligence” is a good description of the place field properties needed in the brainwaves required spatial reasoning network I developed and experiment with, where my primary focus has been where the 3 layer hippocampal tip…

In the video I made you can see “Two Spatial Frames” working together, in each picture showing the network before and after propagating signals at that timestep in time. One frame maps the relevant memories of moving objects into the picture and the other frame maps the stationary objects, which are in this case (other than the food) invisible. It’s not necessary to map out the whole room and all in it at the same time, but in code it’s easiest to include all anyway.

It’s maybe still too early to know how well it models biology, but the average signal ratio matches live animal recordings and the virtual critter does as well or better than a live rat, at a very difficult task. I had to speculate in regards to the network rules that here turned out to only require those of reciprocal connections, to produce a vector map to navigate from.

[Casey]

@mthiboust

Concerning the excitatory cells, I was referring to L5 PT and L6 CT cortical pyramidal cells that project to the thalamus, and to the miscalled thalamic “relay” cells (“miscalled” because they probably do more than a simple relay)

It seems that virtually all axons between the thalamus and the cortex give some collaterals in the TRN (which sends inhibitory inputs to thalamic relay cells and/or other inhibitory thalamic interneurons).

Thanks for the clarification. At least in the rodent somatosensory system, L5 doesn’t form collaterals in the TRN, or so rarely that they’re probably just an artifact of biological processes. All other types of connections between cortex and thalamus form collaterals in the TRN. That includes both upper and lower L6 CT cells, which have different but not mutually exclusive targets in thalamus and TRN.

[Gary_Gaulin]

I was thinking that this quick slide presentation for the “scientific method” might help explain the (HTM-like with ~28 dense SDR bits limit) motor memory system where HTM should work too. Success or failure at a prediction would depend on whether the guessed response motor action brings the head/body direction closer or further away from the (travels upstream) wave direction at that place in the “map”.

ScientificMethod.pdf

For quick responses it’s best to not go too far with resolution. Blobby areas of motion still get us around.

You would know you have the HTM part right by it reversing direction after feeling the confidence destroying zap, instead of like a zombie stop for nothing regardless of how bad and take the shock. There is then a protective part of itself learning how to avoid trouble.

Where shock is made less unpleasant by no longer sensing that bit at every timestep the critter will at some point intentionally endure the discomfort, depending how often the hungry bit chimes in. In the video it’s hungry all the time, but where not it could like go take a nap in the shock free center, then when hungry enough be back in action.

Examples like becoming noticeably agitated when the shock zone is around the food help demonstrate the complex behaviors expected of a real animal, without having to code anything other than a motor HTM system and navigational network to provide directional vector at a given place.

You have to picture a (at least I believe so) grade school easy interaction that (no surprise) has the same features of the revered “scientific method” of human behavior. Where I look for evidence of the “method” in use elsewhere I now end up back millions of years at an earlier than thought possible origin of “scientific thinking” then back in grade school with what I wish my science teachers could have said, to help me figure out what it actually is. The slide presentation would be something I have to put on my wall and spend a year or more trying to figure out the details.

Hopefully the additional information will help make sense of how the two main parts work together.