Evolution & Neuronal connectivity & visual learning.
Evolution & Neuronal connectivity & visual learning.
Nov 07, 2013, 03:47 PM
Group: Basic Member
Joined: Nov 07, 2013
Member No.: 36292
I have become fascinated by the operations of the brain in general, and how the various parts of the brain function together; and am interested in modeling/simulating the evolution of the brain. I ultimately hope (at present, as a learning experiment for another thread) to develop a computer simulation of an evolving brain perhaps as complicated as a rat brain -- based loosely on the morphology and organization of an actual rat or rat ancestor brain.
But here: I'm hoping for an open discussion about the subject of brain morphology/function at the neuronal level and various sources of information, and theories, about how the brain evolved as source material for my project -- ( I'm willing to share my results too, but that means the speculations have to be specific and implementable. )
Some years ago, I attended college to learn about biomedical engineering -- but the field was extremely limited, and so I went into the electrical engineering program instead -- but recently, while doing work on machine vision, I tried looking into human vision to see what new things have been learned in the last 25 years that might be helpful in my projects -- and was dumbfounded by the massive advances in available information in neural-science; It was exciting! I was able to use techniques I learned from engineering on new high resolution pictures of neuron pathways of the eye -- to figure out how to solve several problems I had in machine vision, elegantly. Basically, studying neuronal basis of vision -- provided me a way to build a system which is able to "see" crudely, but better than what I could think up by myself.
I was surprised that an analogy to engineering printed circuit board technology exists in nerve wiring; eg: you would call it "stratification?", which immediately allowed me to reverse engineer *exactly* how the eye was processing the information from the rods and cones without mixing it all up, and I could apply it to an electronic circuit immediately.
However, when I went on to study other parts of the brain -- although I can see the same type of "stratum" theme repeating over and over again, but the detail of the picutres (golgi-staining) etc. is not good enough for me to reverse engineer something as simple as the ear, or even get a good idea of how anything beyond the cochlea itself is connected inside the brain. ( I was able to reconstruct the cochlea, just fine, and compute pitch sensitivity, etc, and get correct answers, but not learn how the brain figures out what is a "harmonic" sound and what isn't. )
The problem is that the micrographs are very limited,and generally only a 2D section is available although the nerve axons and dendrites often spread out in 3D.
So, the first thing I'd love to know -- (probably a vain hope) -- is: has anyone ever done sectioning of a brain, say a frozen one that doesn't distort as sectioned, at successively deeper locations in very small increments; say 1 micron, with photos taken at each layer so that I could reconstruct all the nerve connections in a brain area using a computer program to match up nerve cross sections from successive layers ? If so, who, and where can I get a look at it?
It would be incredibly useful, if someone did have a cross sectioning study ... and ... I'd be willing to share a reconstruction of the nerve connections -- if someone has the tools to make the sectioning.
But assuming this doesn't exist; I'm planning on setting up a simulation of a brain in a computer with what I do know, so far, and seeing if I can program it to evolve into something like an insect or rat brain over time (I'll open that part of the thread in the AI section of this site);
But I need to know what drives the evolution and learning of the brain to make educated guesses as to what will "evolve" efficiently.
I'm wanting a sort of block diagram of how the brain works, with emphasis on how wide the data-path (in number of axons) is between different places in the brain ; for I have developed software which can reliably recreate (synthesize) any logic which the inputs and outputs are known; even if the logic is sequential, hebbian, etc. But it's rate of convergence depends on the efficiency/topology of the circuit -- and I'm hoping that if I can get an idea of a pattern of how nerves are likely to "evolve" over time, that I can avoid wasting millenia on useless guesses....
Take for example the hippocampus, which has a repeating pattern of about five neurons in a looping sequence -- but with unknown lateral connections (at least I only have to guess at one dimension, so it's better than nothing.)
Copying the functionality of that circuit would not be difficult for me; and I know what this particular pattern of neurons is good at:
The hipocampus is good at processing/memorizsing spatial relationship data, and detecting changes in the environment (novelty detection), triggers exploration, and controls learning of the non Stimulus-Response type (non conditioned responses). But the hipocampus is NOT good at shape detection and discrimination.
It can be learned about here (some chapters are a bit boring and dry, so either ask for guidance -- or skim)
h t t p : / / w w w .cognitivemap . net /HCMpdf/HCMChapters.ht ml
But while studying the hippocampus (and learning that I don't need it for a minimal rat -- because a rat lives on, eating and sleeping without it) -- I learned about another block in the rat brain, the "taxonometric" systems ; which a rat can't live without (eg: on it's own, without life support).
But I don't have any information on the topology of the organ which I have to make -- but tons on one that I don't have to make. eg: Where in the brain is/are the taxonometric learnings stored in?
for I know that the brain learns to distinguish objects (taxonomy) but how does it encode the signal(s)?
Is it one nerve output per object, or is it a combination of nerves reacting to attributes of the object which form a "code"?
How can we be certain, and how would I find this out?
In an evolutionary sense, the ability to "see" a line, or an apple, etc: requires a neuronal circuit which is built upon simpler problems solved earlier in evolution, to finally solve the "taxonomy" problem; and one of the things I think is obvious, is that it must be able to reliably produce the *same* code for the same object -- or else everything learned would be in confusion....
So, how does the brain -- which forms stochastically regular networks -- decide to use the same "code" if (for example) the same object is seen in one part of the eye, verses another? eg: unless it is using exactly the same neuron circuit for the same object, then there has to be a way that all the different neurons connecting all the different parts of the eye would agree to the same Code -- they would all have to "learn" to use the same code; BUT: OTOH If the eye is using the same neuron for the same object, then that implies images in different parts of the eye have to be ROUTED to the detector circuit -- and how do they know from a quasi random red green blue placement of cones and rods which ones they should route?
What I'm wanting to focus on is the general idea, how does the eye come to know a straight line/apple/circule is what it is no matter where it is found in the eye? (How did they "learn" it, either evolutionarily -- from simple light detecting nerve, to an eye, or from infant with no knowledge of shape -- to adult with ability to detect shapes in any part of either eye, from either half of a brain...)
Feel free to point me to resources, which I can read AND discuss with you; for I don't mind doing the footwork for a good conversation.
Dec 31, 2013, 10:11 PM
Group: Basic Member
Joined: Dec 31, 2013
Member No.: 36894
Up to 70% of the information that we have of the world comes from our vision, in addition to that we use our vision to verify our other senses.(teachings at an association for the blind where I worked with blind elementary students.) While at university,I was instructed,"When a relatively permanent change of behavior occurs as a result of experience,learning has occured."Back in the day (circa 1977)" we were shown one film that openly demonstrated a mainframe computer conected to (Sonar or radar) that had the capacity to identify objects such as a table or a chair etc. The hope at the time was to give psychologist some insights as to how the mind organizes data.
While working with the blind certain training in occupational therapy revealed that congenitally blind children developed neurologically different than children who were adventiciously blind.(For further enquires and fuller understanding of this information might I suggest Parallel and Divergent patterns of Development in Blind and sighted children between the ages of
3 months and 3 years. I would recomend contacting an association for the blind and picking the brain of their occupational theapist, perapatologist,Special ed staff. There are skills and concepts not easily learned by congenitally blind people the reason being is tha certain behaviors or concepts are best learned during certain critical time periods.After such time the behavior being learned is reffered to as a splinter skill. While trying to deal with teaching students who have splinter skills (If done successfully) an actual change in the wiring of the brain occurs. The closest analogy would be that of an analysan attemting to go back in time (Psychologically) to relive earlier traumas, This time while in the safety of a trusted therapist.
Sources I might suggest an association for the blind,A neurological hospital, some of the works of Wilder Penfield.
|Lo-Fi Version||Time is now: 22nd May 2017 - 10:56 PM|