10 ms·
Key to understanding the thesis: “They found that groups of neurons connect into 'cliques', and that the number of neurons in a clique would lead to its size a
by ianai 7y ago
Key to understanding the thesis:
“They found that groups of neurons connect into 'cliques', and that the number of neurons in a clique would lead to its size as a high-dimensional geometric object (a mathematical dimensional concept, not a space-time one).”
Edit-more to the point;
“"Networks are often analysed in terms of groups of nodes that are all-to-all connected, known as cliques. The number of neurons in a clique determines its size, or more formally, its dimension," the researchers explained in the 2017 paper.”
- Zenst 7y agoMay well turn out later on that AI is best modeled for speed and detail at 11 bit resolution - after all - evolution has done a lot of work that we equate as machine learning. Also interesting that 11 is also a prime number. But equally, may just be one of life's coincidences and the human drive to see patterns and understand exactly what is going on - a trait that still advances humanity today.
- jayd16 7y agoEvolution also determined a giraffe's lyrngeal nerve should detour down the neck and back.
- jacquesm 7y agoApparently the signal arrived too early so a delay loop was spliced in. Score one for ID.
- therein 7y agoWhen the giraffe class was first released it was too OP with its tall neck, allowing it to access food previously unreachable to many other classes while also being able to keep an eye on predators lurking at distance. Upon next evolutionary tick, it had to be nerfed.
- mockingbirdy 7y agoFor anyone who doesn't get the reference: Watch TierZoo [1]. Hilarious and informative, especially for gamers. [1]: https://www.youtube.com/watch?v=6xbQ2WbTp0E https://www.youtube.com/watch?v=6xbQ2WbTp0E
- fucking_tragedy 7y ago> May well turn out later on that AI is best modeled for speed and detail at 11 bit resolution - after all - evolution has done a lot of work that we equate as machine learning. I don't see gradient descent or back propagation in biology.
- Zenst 7y agoMaybe not yet, but for example gradient descent - https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0219004 https://journals.plos.org/plosone/article?id=10.1371/journal... As for back propagation, I'm sure there are aspects in nature in which translate - maybe the nerves system of some species. Way to see it is that evolutionary change of any form could in theory be modeled in machine learning. There may be aspects of evolution and how it operates that help refine how we do machine learning.
- salty_biscuits 7y agoWell for gradient ascent you have this. https://en.wikipedia.org/wiki/Chemotaxis https://en.wikipedia.org/wiki/Chemotaxis
- ianai 7y agoiirc, back propagation sounds similar in inspection to some of the processes seen in sleeping brains. Ie what they see in MRIs looks like what back propagation might look like from outside. But I’m really not comfortable enough with these terms to speak about them.
- DoctorOetker 7y agoThe jury is still out on back-propagation (i.e. reverse accumulation automatic differentiation) within the brain (retrograde signalling from the post-synaptic neuron to the pre-synaptic neuron may still be discovered and turn out to approximate RM AD) But regarding evolution implementing back-propagation, I agree that is just bonkers, as it would require cells to back-propagate in time information about death or procreation! If our biology implements information time machines, biology would be a lot weirder than it is (and it would imply that we could use that by sequencing an organisms genome (or ensemble of them) to read from the future, and keeping it fed without allowing reproduction, and later when we want to write information back either kill or let it reproduce). That's just plain nonsense. There can't be back-propagation in evolution without biological time-travel of information. And machine learning without back-propagation is basically a genetic algorithm, so the claim that "evolution has done a lot of work that we equate as machine learning" would then reduce to a widely accepted fact that nobody disputes and is thus uninformative: "evolution has done a lot of work that we equate as genetic algorithms"
- wbhart 7y agoRight. In other words, it has nothing to do with the human brain creating structures in 11 dimensions. It's a total misreading of the science. This particular claim has been making the rounds for quite some time. It's total rubbish.
- fao_ 7y ago> In other words, it has nothing to do with the human brain creating structures in 11 dimensions. I think you misread that. It means that the title is technically correct but relies on a different definition of the term 'dimension', not that the title is wrong.
- wbhart 7y agoI don't agree. The brain does not create structures in up to 11 dimensions, full stop. This is a popular science article. To use that word from standard vocabulary to describe something that just happens to use the same word as some utterly esoteric mathematical definition is misleading at best. And then to imply that the brain creates such structures is just compounding the problem. What about, "The amazing connectivity of the human brain", or "Researchers unravel the mathematics of connections in the human brain"? There's absolutely no need to confuse readers by conflating unrelated concepts. I could make up my own definition of dimension, e.g. the dimension of a manifold needed to embed some bizarre variety with some cleverly defined homological correspondence to slices of brain material I looked at under the microscope, thought of as simplices. I might then get 276 dimensions. Do I then get a press release for this? It's utterly meaningless. "Mathematician discovers that our brains live in a 276 dimensional universe".
- fao_ 7y ago> The brain does not create structures in up to 11 dimensions, full stop. Except it clearly does, using the definition of the mathematics used to analyse connections. > I could make up my own definition of dimension, e.g. the dimension of a manifold needed to embed some bizarre variety with some cleverly defined homological correspondence to slices of brain material I looked at under the microscope, thought of as simplices. I might then get 276 dimensions. Do I then get a press release for this? It's utterly meaningless Just because we can make up words, all words that have different meanings in different contexts, are meaningless? I agree that perhaps not enough context was provided, but to say that the words itself do not have a certain meaning because you misunderstood the context in which it's used is ludicrous. The world does not revolve around you.
- karmakaze 7y agoIt's hard to tell from the post what the terms are being used to describe. First of all, I would consider a dimension either space-time or mathematical to be an orthogonal/independent concern. Now where this gets distractingly blurry is that the paper itself says "The number of neurons in a clique determines its size, or more formally, its dimension." which uses dimension to mean size, like a length. If I'm generous I could maybe believe that larger cliques form more complex connection patterns and can perform higher dimensional processing. But even this fails as the paper then goes on to describe the cliques and cavities as being higher-dimensional. All that aside, I do think there's something to the paper, just using some poor choice of terms. It seems what's being described as complex is both the physical non-uniform pattern of connection and the subgraphs of activation which may be able to consider a high number of dimensions for a given input stimulus. I couldn't follow the Simplices/Complexes/Betti numbers to see how it relates to a measure of dimension.
- Mirioron 7y agoWhy would the space-time and mathematical concept of dimensions be orthogonal? Don't we use the mathematical concept of dimension when dealing with space-time dimensions? Software developers in particular should have intuitive understanding of this connection, because implementing space-time in code is usually done through methods that closely resemble mathematical dimensions. Eg if you were to implement a point in 3 dimensional space time you'd give the point xyz and t as variables.
- karmakaze 7y agoI don't mean the math one orthogonal to the space-time one. I mean that dimensions are orthogonal to each other whether in math or space-time (except special relativities).
- eli_gottlieb 7y agoOh ok, so they actually mean that they found 11-dimensional directed simplices in neuronal connectivity, which then implies the barycentric coordinates describing relative activity levels in the simplex will have 11 dimensions. Simple.
- zitterbewegung 7y agoThe correct title to this article is that the topology of the Human brain corresponds to a set of connections in an eleven dimensional space.
- respinal 7y agoStill too long to post it here..