Mapping the Brain's Neural Network

Ponca City, We Love You writes "New technologies could soon allow scientists to generate a complete wiring diagram of a piece of brain. With an estimated 100 billion neurons and 100 trillion synapses in the human brain, creating an all-encompassing map of even a small chunk is a daunting task. Only one organism's complete wiring diagram now exists: that of the microscopic worm C. elegans, which contains a mere 302 neurons. The C. elegans mapping effort took more than a decade to complete. Research teams at MIT and at Heidelberg in Germany are experimenting with different approaches to speed up the process of mapping neural connections. The Germans start with a small block of brain tissue and bounce electrons off the top of the block to generate a cross-sectional picture of the nerve fibers. They then take a very thin slice, 30 nanometers, off the top of the block. 'Repeat this [process] thousands of times, and you can make your way through maybe the whole fly brain,' says the lead researcher. They are training an artificial neural network to emulate the human process of tracing neural connections to speed the process about 100- to 1000-fold. They estimate that they need a further factor of a million to analyze useful chunks of the human brain in reasonable times."

Missing tag

[2.7182 (819680)]

Vaporbiology

Re:

[Feminist-Mom (816033)]

Not to mention - and I a sure I'll get modded down here - is that neural networks aren't very effective. They get a lot of hype in the media, and people who don't work in optimization like them. They sound appealing and cool, but there are other methods that are much better. If you look at the liturature, there is not that much hard science behind them. Some statistical mechanics people have some results, but they really are just a fad, like fractals or catastrophe theory. But bear in mind LOTS of people i

Feedforward NNs versus biological NNs

[benhocking (724439)]

Many engineers have studies feed-forward neural networks and found them to be far inferior to other solutions. Of course, our brains use recurrent neural networks, which, unfortunately for engineers, are very difficult to analyze. There are many secrets yet to be teased out of these neural networks, but much progress has already been made. Researchers in our lab, for example, have demonstrated how introducing random synaptic failures improves not only energy efficiencies, but also the cognitive abilities of simulated neural networks. I'm currently researching the effects of variable activity (as measured in a biological neural network by an EEG), and I dare say there's a lot more that we don't know about these networks than we do.

Recurrent neural networks

[benhocking (724439)]

I'm writing my doctoral dissertation on the CA3 region of the hippocampus, although this is also true for the entire brain. In short, I've done a lot of reading on how the brain is wired in an attempt to reproduce it in silico. There's still a lot for us to learn about how the brain has accomplished certain feats.

Re:Missing tag

[Space cowboy (13680)]

Perhaps you didn't get much out of neural networks, but my PhD thesis was on the similarities between a kohonen network [wikipedia.org] and relaxation-labelling [ic.ac.uk] equations. Part of it is up on my blog [gornall.net] (I haven't actually got as far as that bit yet, but the groundwork is there).

A neural network (well, anything more complex than the single-layer perceptron [wikipedia.org] anyway) is an arbitrary classifier. I'm curious as to why other methods are "much better". Unless you do an exhaustive search of the feature-space, all classifier methods are subject to the same limitations - local maxima/minima (depending on the algorithm), noise effects, and data dependencies. All of the various algorithms have strengths and weaknesses - in pattern recognition (my field) NN's are pretty darn good actually.

It's also a bit odd to just say 'neural networks' - there are many many variants of network, from Kohonen nets through multi-layer perceptrons, but focussing on the most common (MLP's), there's a huge amount of variation (Radial-basis function networks, real/imaginary space networks, hyperbolic tangent networks, bulk-synchronous parallel error correction networks, error-diffusion networks to name some off the top of my head), and many ways of training all these (back-prop, quick-prop, hyper-prop, batch-error-update, etc. etc.) I guess my point is that you're tarring a large branch of classification science with a very broad brush, at least IMHO.

Not to mention that this is all the single-network stuff. It gets especially interesting when you start modelling networks of networks, and using secondary feature-spaces rather than primary (direct from the image) features. Another part of my thesis was these "context" features - so you can extract a region of interest, determine the features to use to characterise that region, do the same thing for surrounding regions, and present a (primary) network with the primary region features while simultaneously(*) presenting other (secondary) networks with the features for these surrounding regions and feeding the secondary network results in at the same time as the primary network gets its raw feature data. This is a similar concept (if different implementation) to the eye's centre-surround pattern, and works very well.

If you work through the maths, there's no real difference between a large network and a network of networks, but the training-time is significantly less (and the fitness landscape is smoother), so in practice the results are better, even if in theory they ought to be the same. I was using techniques like these almost 20 years ago, and still (very successfully, I might add) use neural networks today. If it's a fad, it's a relatively long-running one.

Simon.

(*) In practice, you time-offset the secondary network processing from the primary network, so the results of the secondary networks are available when the primary network runs. Since we still run primarily-serial computers, the parallelism isn't there to run all of these simultaneously. This is just an implementation detail though...

Re:Missing tag

[mesterha (110796)]

A neural network (well, anything more complex than the single-layer perceptron [wikipedia.org] anyway) is an arbitrary classifier. I'm curious as to why other methods are "much better". Unless you do an exhaustive search of the feature-space, all classifier methods are subject to the same limitations - local maxima/minima (depending on the algorithm), noise effects, and data dependencies. All of the various algorithms have strengths and weaknesses - in pattern recognition (my field) NN's are pretty darn good actually.

Actually, more recent methods don't have local maxima/minima. Something like a support vector machine optimizes an objective function. Of course, this is somewhat of a tangent, in that the objective function might not be a useful metric for performance, but people have shown that the minimum objective function value of a SVM does relate to its generalization performance. It's a little disconcerting that a NN has an objective function but that it can find it's minimum or that the minimum doesn't give good performance on test data (over-fitting)...

Of course, part of the NN's problem stems from the fact that it is an arbitrary classifier. It's hard to give generalization results for an algorithm that has an infinite VC dimension. (There are techniques to restrict the size of the weights to give some guarantees.) However, this doesn't mean NNs can't perform well in practice. It probably means that the current theoretical analysis is somewhat flawed in relation to the real world.

So have you ever compared your NN algorithms with the popular algorithms of the day such as SVMs with kernels or boosting algorithms. Also, are your NN algorithms generic or do you heavily customize and tweak to get good performance.

Re:

[Space cowboy (13680)]

As far as generalisation goes, with the stuff I do there is generally an abundance of training data - it's easy to get imagery, and it's easy to generate the input data from that imagery. As such, using some of the data as training data, and the remainder as generalisation-test data is a good (and pretty standard) approach. All you do is train on (say) 2/3 of the data while simultaneously checking the results against the remaining 1/3 of the data. You'll get two error-curves, which initially approximate eac

Re:

[schnipschnap (739127)]

The pig go. Go is to the fountain. The pig put foot. Grunt. Foot in what? ketchup. The dove fly. Fly is in sky. The dove drop something. The something on the pig. The pig disgusting. The pig rattle. Rattle with dove. The dove angry. The pig leave. The dove produce. Produce is chicken wing. With wing bark. No Quack.

Re:

[mi (197448)]

They sound appealing and cool, but there are other methods that are much better.

Much better for what? If, in 30 years, this technology allows full mapping of the entire human brain, I'll be quite happy... It will mean, that my conscience may live after the mortal flesh dies.

It may take another hundred years or more for full reconstruction of a new brain (and the rest of the body) to become possible, but in the mean time I'll be preserved just as I was, when I died and my brain was scanned.

Even better,

No GPS Needed

[pipingguy (566974)]

In my case it'd be trivial: brain cell 1 ---> brain cell 2. Done!

Did I get that right?

[UbuntuDupe (970646)]

They are training an artificial neural network to emulate the human process of tracing neural connections to speed the process about 100- to 1000-fold.

So, they're training a neural network to automate the process of mapping a neural network, in the hopes of creating an intelligence that they can train to automate other processes?

My brain hurts...

Re:

[Air-conditioned cowh (552882)]

"My brain hurts..." Please hold still. It's being mapped. Thanks.

What constitutes a "map" here differs from elegans

[jpfed (1095443)]

One reason we could do the aforementioned mapping with C. elegans is that the worm's neurons are always laid out the same way from worm to worm. This is not the case for humans, and probably not the case for any vertebrate.

Interesting

[rm999 (775449)]

Sounds interesting, but is a map enough to understand the brain? I know in artificial neural networks, the actual structure isn't as important as the weights on the nodes. Will hitting the brain with electrons be enough to give us an understanding of these "weights", or just the connections between them?

Horse, push cart.

[zippthorne (748122)]

Science is all about the baby steps. You can't talk about determining the weights before you know what the connections are.

Re:

[bazald (886779)]

Or more to the point, is a neural network really the correct interpretation of a brain's structure? I suppose we might find out soon, if they can get the "weights" "right".

Re:

[melted (227442)]

Weights are trained and they'll be different depending on the connections and on the training. The rough structure of long-range connections is actually fundamental to how the brain works. This is essential to explain long and short term memory, fear, involuntary reactions to stimuli, perception of the world as a coherent audio/visual picture as opposed to a set of disparate inputs, etc.

Re:

[unixfan (571579)]

Hell, it will be fun to see what they figure out to be the capacity of the brain, as far as how much information it can store.

I know visually we are looking at least at something around 24 frames per second. The eye is supposed to have a resolution of around 1000 dpi. Not sure how to measure the viewing area. But let's say it is lesser and lesser resolution the higher the angle. Let's say, just to have a number, that we have a 16:9 viewing ratio at two feet distance. Lets say it's three feet wide. That shou

Your post are so full of bullshit..

[aliquis (678370)]

.. I don't know where to start.

Repeat after me:
YOUR EYES AND BRAIN AREN'T DIGITAL FINALSPEC1.0 DESIGN.

MOVIES use 24 fps with motion blur because it gets an acceptable quality, your eyes can still notice things happing during 1/200 of a second.
In our digital graphics class the teacher mentioned something like 20 million cells in the eyes which registred data but only one million nervs to the brain.
YOU HAVE SHIT POOR QUALITY OUTSIDE ONLY A FEW PERCENT OF THE CENTRUM OF YOUR VISION.
I guess your dpi esimate may

Re:Interesting

[smallfries (601545)]

When you pull figures from your arse, you don't actually add anything to the discussion. All that we've learned is that your arse is rather large, and you are used to removing things from it.

Even if your (incorrect assumptions) were correct, 36" x 20" at 1000dpi would be 36000 pixels x 20000 pixels = 720M pixels. Clue: dpi is a scalar measure rather than area.

Of course, the human eye does not work anything like that. Rather than farting numbers I spent 10 seconds on Google to find this [ndt-ed.org] which looks into the question of Visual Acuity. The "high-res" part of the eye is a very small circle with about 120 "dots" across its diameter.

As we do not resolve entire "frames" in a single go, the concept of a frame-rate is completely ludicrous. Your argument earlier in the thread about observing skipping when seeing a high speed stimuli doesn't show evidence of a *periodic* frame rate. It just shows that there is a *minimum* temporal resolution. One does not imply the other, especially when the eye is processing asychronous input (from rods and cones).

Although you don't believe that the brain fills in the missing images with educated guesswork, we've already established that what you believe is shit. Most (if not all) neuroscientists have accepted that the high resolution continuous visual imagery that we see is mostly an illusion produced by the mind. There are many well reported experiments that provide evidence of this. You should look for anything on Visual Illusions - there are far too many decent results in peer reviewed journals for me to spend time looking for you. Change Blindness is a related phenomena.

Finally you've cooked up some stupid figures for the number of cells in a brain. Why do you feel the need to demonstrate how stupid you are? The actual numbers (which you get wrong by 3 fucking orders of magnitude) are in the summary of the article! How hard is it to read the 100 billion neurons at the top of the page.

So next time you feel the need to pontificate needless about something that you don't know anything about. Don't. You, sir, are a thief of oxygen and your pointless ramblings have made everyone reading this article collectively dumber.

PS Feel free to mod me flamebait, as I am clearly annoyed. But when you do so remember that the everything the parent poster wrote was incorrect and that I have pointed out to him where he is wrong.

Re:

[smallfries (601545)]

We both agree that your numbers weakened your point as they were clearly wrong.

While I don't practice neurobiology myself, my girlfriend's PhD was in psycho-physics and how to exploit the "compression" in the human visual system. Her research was very much at the practical end of the field.

Where we disagree is on whether or not your point stands. You claimed that despite the lack of bandwidth between the eye and the brain, the brain was *not* responsible for synthesizing the majority of what we think that w

Re:

[steelfood (895457)]

A) The eye compresses images even before it gets sent to the brain.

B) We don't memorize the entirety of whole scenes (not even those with photographic memories, though they're close). We use pattern recognition. That's why you can tell coke can is a coke can given from a slightly dented coke can to a crushed coke can, and can tell that even a crushed coke can isn't a crushed coke bottle. You memorize the patterns that make up a particular concept, match any given object to your set of memorized patterns (re

You are a simmulant

[goombah99 (560566)]

Bad news. You are a simulation on a playstation 9, not a real human.

No really. it's overwhelmingly probable you are a simulation.

According to the article above that are 100 trillion neurons to simulate. Even if they were multi-state that's approaching trivial by computational standards. And if you are willing to run the simulation at sub real time you could do it now.

So according to the anthropic principle, either 1) the human race goes extinct in the near future before this level of simmulation is poss

Re:

[no-body (127863)]

Will hitting the brain with electrons be enough to give us an understanding of these "weights", or just the connections between them?

Nope - it will be just the wire connections, just as if computer hardware without any software will have it's electrical circuits traced in order to understand better what a program running on the screen does - about in that magnitude, probaly much higher.

The "software" on a human brain is programmed from before birth and constantly changed.
Just the computing power of keep

Re:

[yet another coward (510)]

RTFA, rm999 and moderators. This very point comes up at the end.

Computational Complexity

[Yahma (1004476)]

It's quite interesting that these German researchers are mapping pieces of the brain; however, even if they were to map the entire human brain, first, we still do not know how to perfectly simulation the biological processes occuring in the brain. Yes, we are able to simulate a single neuron, or small clumps of neurons; however, the dynamics of simulating billions of interconnected neurons is not fully understood.

Second, even if we were able to map the entire human brain and run a perfect simulation, the

Re:

[Renraku (518261)]

Lets say that we could, in fact, emulate an entire human brain in real time.

It would be very difficult to get something useful out of it. Answers wouldn't always be the same due to the semi-random effect brains have a tendency to produce. You would spend a lifetime putting something in and watching it come back differently than it did a few minutes ago. It wouldn't be too unlike a network connection, if packets were voltage gradients and various neurotransmitters. Since there are only a few ways each ca

Re:

[Fourier404 (1129107)]

If you emulate a human brain in real time, as well as connections to emulated eyes, ears, and mouths, you just have to talk to it for a couple years and it'll learn English. =D

Re:

[Renraku (518261)]

First you'd have to have the IO systems connected properly. If you just grow a brain, this doesn't happen. For all practical purposes, the spine is part of the brain. The optic nerves are part of the brain. The cranial nerves are brain too. They all grow out of the brain when the brain is given chemical signals at the right time. Or more specifically, the stem cells are. If its not given, they don't grow. The brain will be a closed system.

The best idea for that would be to get the brain, cranial ner

Re:

[bmo (77928)]

"It would be very difficult to get something useful out of it. Answers wouldn't always be the same due to the semi-random effect brains have a tendency to produce. You would spend a lifetime putting something in and watching it come back differently than it did a few minutes ago."

We call this "raising children"

--
BMO

Re:

[bmo (77928)]

Yeah, unfortunately I got modded "redundant" and "funny." I also got modded "redundant" for my Beck joke. How is it redundant when I was the only one to come up with that? What...ever. I got more karma than I know what to do with. Go ahead, waste your mod points on me.

If we truely come up with computers that mimic the human brain, we're going to see the same problems that we have "programming" children, maybe even worse, because human children have inherited behaviors that make teaching easier, and elec

Re:

[rbarreira (836272)]

It may not be possible to do so without changing the behavior of the cognitive part. Investigating if that was the case or not would already be an extremely important milestone!

CCortex

[Colin Smith (2679)]

http://www.ad.com/ [ad.com]

An attempt to emulate a brain on a network of computers.
 

Plasticity

[DynaSoar (714234)]

That's what they call the brain's ability to change. By the time they complete a wiring diagram, it'll have changed. Also, knowing the wiring and connections is not enough. Knowing which connections are excitatory and which are inhibitory is necessary, and then tracking down loops of excitatory against excitatory resulting in inhibitory, etc. It's all fine and well to have a map, but that doesn't tell you squat about what anything does. A useful map would have to be dynamic, and the complexity of that is far more than just what they're considering for a wiring diagram.

they got the complete neural map of C.Elegans

[Z80a (971949)]

and it's only like 302 neurons,so,it's possible to write a simulator of it?

Re:they got the complete neural map of C.Elegans

[lazybratsche (947030)]

Well, that's the hope, and a major source of appeal for the humble nematode. Unfortunately, that's still far beyond what we know right now. The physical map of every neuron and their connections has been complete for decades. Still, despite a whole lot of effort, researchers are still working to piece together small functional circuits for the simplest of behaviors. A lot of complexity arises in neural circuits -- one physical circuit can contain several independent functional circuits, depending on the types of inputs.

The best current knowledge of C. elegans neurophysiology involves qualitative descriptions of small circuits, involving a few dozen neurons. Unfortunately, while you can do a lot of good behavioral studies and other experiments, it's impossible to directly record the activity of specific neurons. Also, it turns out that some "neural" functions are actually performed by other cells. For example, one pattern generator in the digestive tract actually resides in intestinal cells instead of neurons -- my lab is working on the genetics involved.

This shit gets complicated, fast.

IAAUCER
I am an (undergrad) C. Elegans researcher

Re:

[Tablizer (95088)]

[worm has] only like 302 neurons,so,it's possible to write a simulator of it?

while (! dead) {
    if (leftTenticalSensesFood()){wiggle(left);}
    if (rightTenticalSensesFood()){wiggle(right);}
    if (frontTenticalSensesFood()){munch();}
    if (femaleWormEncountered()){fuckTheWigglyMamma();}
}
end();

what does a neuron map look like?

[wikinerd (809585)]

The neuron map of the average human brain looks like this [venganza.org]. More evidence for the existence of him [venganza.org]!

100 trillion synapses?

[SolusSD (680489)]

That's simply not true. High estimates for the total number of cells in the human body are 100 trillion, a quick google search yields this: "Science NetLinks, a resource for science teachers, stated that there are approximately "ten to the 14th power" (that's 100 trillion) cells in the human body." maybe they meant 100 billion? I remember from one of my AI classes that there are up to 10,000 synapses per neuron, while other have fewer than 100.

Oblig. Futurama

[Dachannien (617929)]

Farnsworth: Lie down here and we'll do some tests. If Fry is out there, then Leela's brain could be acting as a five-pound Ouija board.
Leela: Is this some sort of brain scanner?
Farnsworth: Some sort, yes. In France, it's called a guillotine.
Leela: Professor! Can't you examine my brain without removing it?
Farnsworth: Yes, easily!

Me!

[SEWilco (27983)]

Map my brain!
...reads article...
When I'm done with it!

Perhaps

[reboot246 (623534)]

they should start with an easier project. What's CowboyNeal doing for the next few days?

Re:

[reboot246 (623534)]

Who cares?

Maybe it's not all that bad ...

[ScrewMaster (602015)]

keep in mind that huge swaths of the human brain are for data storage, motor control, sensory processing, autonomic functions and other elements not directly related to sentience. If your intent is to use this information to produce a viable synthetic human intelligence, you wouldn't necessarily need to model the who shebang.

Yes, but ...

[PPH (736903)]

... when the map is done, will anyone ever be able to re-fold it?

Bruce McCormick

[eric76 (679787)]

A professor at Texas A&M University, Bruce McCormick, was pushing for this for years.

Check out Welcome to the Brain Networks Laboratory at Texas A&M University! [tamu.edu].

The idea is to use a knife-edge scanning microscope to make images of very thing slices from brains.

I'm curious if Dr. McCormick has retired. His web page last list courses he taught in 2002.

Re:

[canuck57 (662392)]

Our computer technologies have yet to achieve the complexity of most biological brains. I'd love to see these new informations derive a new form of super-computer. Of course...We have to watch out for iRobot scenarios...

Don't hold your breath for an iRobot.

If each of the 100 billion neurons managed the 1000 or so synapses, and say a modern day PC with a quad processor could computationally handle say 100 neurons, you would need 1 billion PCs. Since 1 billion PCs would find it difficult to walk, the old

Re:

[multi io (640409)]

Now lets say usable processing power doubles every 5 years, and it shrinks to something small enough it can walk into our living rooms. That would be at least 150 years from now.

That's only as long as you simulate neurons in software, which is probably as inefficient as it gets, as opposed to building and connecting artificial neurons in hardware directly. The fact that the brain manages to cramp the intellectual differences between reptiles and humans into a few cubic centimeters should tell you some

Re:

[Metasquares (555685)]

Ah, but can such an approximation truly result in intelligence or would it just be a very slow and complex classifier? Or both? :)

Are all of the functions of a neuron required to produce intelligent behavior? If not, which can we omit? How will we even know when a system is behaving intelligently? Even humans take years to learn how to communicate and rationalize. Could we provide even a perfect simulation of the human brain the proper environment to train in to ensure these results? Once we have such a mod

Re:

[mikael (484)]

The Halting Problem [wikipedia.org]