Numenta Research Meeting - Dec 9, 2019

These higher level organizations of information is where the multi-map communications come into play.
It is not all done on a single 2D map. Binorality (phase delay) , timbre, pitch, volume, sequence, and rhythm are all features that are to be extracted at a lower level and recombined for higher level pattern extraction.
If you look at the Auditory Cortex you can see that there are many intermap connections.

This follow the general pattern in the visual chain where “simple” features are extracted close to the sensory patches, and progressively more complex combinations of these features can be found as you ascend the hierarchy. As these features become more abstract it gets difficult to relate back to the original stimulus as the response patterns no longer seems to be directly related to the sensory stream where the sensation are received; the inter-dimensional mapping is not clear.

The abstract codes the brain uses do not seem to be organized the same way as the ones we use to encode things in the sensory realm. It is in some high-dimensional space where humans have no direct experience or intuitive understanding. This property defies our usual attempts at symbol decomposition into content.

This is why understanding the theoretical underpinnings of the H of HTM is so important.

I personally feel that the theoretical tools to understanding these high-dimensional codes this will be found in some extension of set theory.

2 Likes

I agree.

But, the problem is way before the H. How the relationship between close mini-columns is established. In other words, what is the purpose of L6->L4 modulatory inputs? Is some form of “spatial” reference? or its just sensory context? (i,e, the representation of close sensors activity).

How do Grid Cells […] get anchored to sensed objects?

The paper (Kropff & Treves, 2008) has a good and biologically constrained theory of how this might happen.

https://onlinelibrary.wiley.com/doi/abs/10.1002/hipo.20520

* I highly recommend this paper. Don’t let the title of it discourage you; the title is a pun, a joke which you will understand after reading it.

3 Likes

This is the heart of both my hex-grid concept and Numenta’s thousand brains theory.
The assumption in both cases is a co-operative voting method to join local fragments of representation into recognition of larger spatial/temporal pattern.

This action is bidirectional so that the joint action acts as a filter or de-ambiguator.

Applied to a larger section of cortex this filtering action acts to both pull signals out of noise and generalize.

It seems not applicable in 3D, according to experimental pieces of evidence.

From https://www.youtube.com/watch?v=QfW5wbSBoKU

this is a related discussion regarding the 2-D vs 3-D nature of grid cell representation here:

I am not really surprised that an inherently 2D sheet preferentially maps 2D space.

1 Like

The fact that grid cells don’t map a perfect lattice is not restricted to 3D.
We can see similar differences in 2D environments after the removal of a boundary.

image
From https://www.nature.com/articles/s41593-017-0036-6

2 Likes

Right. But what if grid cells are in fact encoding “a place” of that “2D sheet”? In other words, you are just translating cortex-addresses to the hippocampus-addresses. You need it to replay later hc the content in the “right” place of the cortex.

I think, we had this discussion before, :slight_smile:

Well actually, I think that evidence supports the (Kropff & Treves, 2008) theory, which explains how grid cells form receptive fields with a “Characteristic Distance”.

Grid fields can change size and deform when a familiar environment changes size or shape.
Experience-dependent rescaling of entorhinal grids

POm, the direct sensory relay to L5a of barrel cortex, has larger RFs and poorly characterized responses compared to VPM, so it fits sending a movement direction signal. The same TC cells are also the CTC pathway though.

L6 CC cells mostly fire at the start of excitation. Maybe that’s a way to encourage activity to move around in the minicolumn. Different rates of adaptation could create a fixed order of activity moving around the minicolumn.

One study found L6 CC cells can also have long latencies before they fire. If the ones with longer latencies in response to weak input also adapt more quickly in response to strong input, they fire first when input is strong and last when input is weak. That way, activity moves around in one order when input is strong (movement in the preferred direction) and the opposite when it’s weak (movement in the opposite direction).
This wouldn’t work for long periods of time. Maybe oscillations solve that, resetting the process each cycle to form 1-d grid fields.

L6 CC cells in barrel cortex receive a very strong input from M1, maybe the strongest in all sublayers.

2 Likes

Sounds interesting, but I don’t get your link with grid cells.

Have you read this paper? L6 CC cells (excited by TC cells) gate L5 PT cells output.
https://www.cell.com/neuron/pdfExtended/S0896-6273(19)30884-0

The paper is not related to grid cells, but it could help with the role of L6 CC cells.

2 Likes

I’m not sure how the gridiness would work, now that I think about it more. My point was more that it doesn’t need to work over long periods of time because the response is gridy. It gets between equivalent points on the 1-d grid before it runs out of cells which still haven’t yet fully adapted. Cells recover from adaptation between each high-activity part of the oscillation.

No. I’ll skim it now.

During the research meeting on dec 9th 2019, @jhawkins talked about two types of sensors:

@subutai confirmed this a bit later.

Does anyone know of a neuroscience paper that discusses these two types? I tried googling for info on this, but I didn’t get far. I don’t really know how to call them or how to descibe this.

2 Likes

Start here:

You have the entire body doing inverse kinematic transform to support the gyroscope stabilization of you visual system. All the way to your supporting surface. Add to that your head and eye pointing system and you have a very good directional system.
Let me go a step further - this same system is your map back to the frame of reference “out there” and your interface between all this and your “end effectors” makes your “personal space” special as it is where your sense of agency has the greatest influence.
This whole thing is mixed in with your posture control system. If you think about it - it is really the same system.

This IS an insanely complicated system and goes directly to the point that the primary function of the brain is to run the body.
This also goes a long way towards explaining this:

Mark, this is all interesting, but where does it describe two types of sensors on our skin or retinas?

No! These senses are part of the vast somato-sensory system.
Joint angles and vestibular sensations are distinct and separate from skin and eye pattern sensations.
I could imagine someone thinking that the location of the pattern in the sensory field is the relative location but I consider that as part of the pattern.

I could be looking at this all wrong but I think that these strained efforts to put location and pattern into the single column is mostly due to a hard focus on the macro column to the exclusion of hierarchy.

I don’t think I understand. Do you mean the same physical sensor in the skin projects a signal to two different centers or regions, where one is interpreted as a fixed spacial location and another is converted into directional information?

Or is it that you disagree with the principle of having two types of sensors?

Or maybe we’re not talking about the same thing here.

This one!

1 Like

I may be off the mark, but I interpreted this comment as simply calling out two different types of nerves. Those which detect movement, and those which detect pressure (I imagine there are many more than that, such as those which detect heat, those which detect the color red, etc. etc.). The interesting thing is that Jeff mentions they travel in separate channels to the cortex, though… not sure what to make of that.

2 Likes