Rodent Neural Activity Has a Geometric Structure (forbes.com)

TheAlexKnapp writes: In a recent paper (abstract), a team used techniques from computational topology to look at the neural activity in the rat hippocampus as it solved a maze. Mathematician Kevin Knudson explains the findings: "This is the first time geometric structure has been found intrinsically in neural data. Certainly such a structure is to be expected since the rat's place cells keep track of the geometry of the environment, but this result is confirmation that it can be detected using only the pattern of correlations among the neurons. And it suggests that such geometric structure is a property of the underlying place cell network and not a result of the spatial structure of the input cells."

Comments

Which kind of rodents?

[nospam007]

Rodents Of Unusual Size? (Rous)

Re:Which kind of rodents?

[avgjoe62]

I don't believe they exist.

Re:Which kind of rodents?

[PolygamousRanchKid]

Rodents Of Unusual Size? (Rous)

Also known as lawyers.

It would be interesting to drop a few lawyers in a maze, and take a peak at their hippocampi . . .

Re:Which kind of rodents?

[Chris Mattern]

Given that we're talking about geometric structures, maybe they're rodents of unusual shape...

Re:Which kind of rodents?

[FreeUser]

Given that we're talking about geometric structures, maybe they're rodents of unusual shape...

The many angled ones. They live at the bottom of the Mandelbrot set. Be afraid. Be very afraid.

Re: Which kind of rodents?

[Anonymous Coward]

-1 Nightmare inducing

Of course

[Anonymous Coward]

It's not just neural activity. All of embryologic development is geometrically mediated. See Erich Blechschmidt.

Explain this please:

[Anonymous Coward]

Certainly such a structure is to be expected since the rat's place cells keep track of the geometry of the environment

What?

Re:Explain this please:

[roman_mir]

I read TFA and the way I understand it is that neurons in the hippocampus connect with each other in a way that resembles or projects a map of the environment into the connected structure. So basically a picture or map of the environment is formed in the brain. A mouse is released into a maze, the maze is a geometrical structure that mouse follows to try and escape. As the mouse follows the structure the picture of the maze is mapped into the brain and stays there for some time (unknown?) The mouse knows where it is in the maze because of that shape of the connected neurons that forms in its hippocampus.

So we probably do the same. The size of the brain (number of neurons and connections between them) likely limits how much of the space around us we can map into it. It would be interesting to find out what is the size of the maze supposed to be before a mouse can no longer figure out where it is? Also what if the maze is dynamically changing (corridors are closed and others are opened after the mouse passes them), how does the brain remap this information, does it add new information without removing old data?

What other meta-data is stored in the brain beside the map of it, how about colours, smells, environment (sharp corners, sticky floors, puddles of water, etc.) How much can a mouse be pushed before it gives up?

Re: Explain this please:

[Anonymous Coward]

The rats brain cells mimic the features of a scene in structure directly rather than some other methods, like some burst of activity without any structure, which is what would be expected under older neurology knowledge.
But the more we look at how the brain is structured (especially just the past few years), the more we realize how wrong we were. The brains layout is highly structured and would make any sysadmins network layout look like a childs Lego construction.
That recent-ish deep scan of neuron structure completely rewrote the text-books. It is amazing how well it has developed, nice parallel lines, smooth curves, even over-under stitch-like structures.

Re:Explain this please:

[Anonymous Coward]

I think what's especially interesting about this is that it sheds further light on how differently brains behave from artificial neural networks. A previous example showing this: The flaw lurking in every deep neural net. In case you don't immediately see the relevance, read this comment in the article discussion. And we already know that there are also parts of the brain that map out numbers and such. Biological neural networks seem to get a kind of structure that is very unlike what you tend to find in artificial networks, something which you'd think would be a very interesting topic in AI research and could lead to major improvements. But even though I follow the field, I don't hear much chatter about this.

Re:Explain this please:

[raftpeople]

"Place" cells track when the animal is in a specific place (relative to past experience and environment), "Grid" cells are physically arranged in a grid (approx equidistant from each other) and fire according to animals location within the current environment (meaning if it's approx in the center of the room, then the approx center cell is firing, if moving north then the sequence of firing cells follows that same path, like an internal grid/map).

Proof of God's design!

[AndyKron]

Proof of God's design! How could that structure just evolve from random mutations? If you saw a geometric pattern in the sand by the beach you would know it was designed by someone. If you saw a Boeing 747 you would know it wasn't made by a tornado in a scrap yard. Checkmate atheist evolutionist, neo-Darwinist liberal Nazi fascists with slight leanings toward Maoism! /s

Re:Proof of God's design!

[I4ko]

this sounds about right

Re:Proof of God's design!

[umafuckit]

You laugh, but I reckon the best argument against the bullshit is just to teach and show them how a fertilized egg develops into an organism. This is the same change in complexity that you see over evolutionary time: where single celled organisms eventually turned into amazing macroscopic things like ducks, trees, and elephants. However, with development you can watch the changes happen in real time and the final result is often obtained in a matter of weeks. Best of all, we're learning a heck of a lot about how it actually happens. So it's not some black box, but an explicable phenomenon.

Re:Proof of God's design!

[Anonymous Coward]

Stories definitely do exist and are passed along the generations. Those gods, they are all networks of interacting stories and always have been. Religions are the ways we implement those stories in our societies and cultures. Stories make us human in all the human and inhumane ways possible.

hyperbolic

[Anonymous Coward]

It's hyperbolic. What, you expected spherical geometry?

Mousetagon

[Anonymous Coward]

2 mice enter
1 mouse leaves

I got a job for you

[future assassin]

it pays good and it also involves 72 virgins.

Grid Cells

[LionKimbro]

I thought we already knew that. Grid Cells. I first learned about them in a 2007 Scientific American Mind issue.

I think the new thing in the article is this particular way of searching for their signature, or something.

more cover stories

[strstr]

old info, being fed to the public that's been denied to them for decades.

drrobertduncan.com/

Symmetric matrix???

[Anonymous Coward]

They found correlations between neurons in a mouse brain as it was solving a maze.
(Maybe each neuron corresponds to a place and the places are somehow connected.)

They have a nxn matrix describing the connectivity between n neurons.
The matrix is symmetric.

I wonder if neurons work with symmetric coupling.
(If A fires, is it just as likely to cause B to fire as the reverse?)

I thought all the low level research as to how neurons were connected showed directionality in the connections.
Which may say their matrix should not be symmetrical?

Otherwise, neat stuff.

A rat hippocampus ran a maze?

[idontgno]

Pray tell, where was the rest of the rat?

Surgically extracting a portion of a brain, giving it senses and mobility, and teaching it to navigate a maze is much more impressive than just discovering that its neural activation is symmetrical.

Try a different maze

[Anonymous Coward]

One maze with rounded corners, that would be fair and square.

clique Homology Betti number?

[Anonymous Coward]

TFA. Is it really impossible to explain this without jumping into the matrix math and jargon? Come on Kevin Knudson, you can do better.

Let me guess...

[hyades1]

The structure resembled a rat trap-ezoid.