College freshman, interested in theoretical neuroscience

[feuxfollets]

So I'm currently a freshman at UPenn. I came in wanting to major in physics & math and go on to grad school in particle physics. However I've been a bit disillusioned since, by how esoteric the math required for it was and by the fact that there are barely any positions in academia for it. And my physics lab TA just finished his PhD on something about Calabi-Yau manifolds and went off to work for an investment bank.

I've also been interested in computer science, specifically artificial intelligence, and this has led to my present interest in theoretical neuroscience. I have a few questions about this field and what would be the best preparation for it.

I've heard that a lot of condensed matter physicists go into this field. Are neurons actually modeled based on fundamental physics? Because the other thread on this forum, as well as the theoretical neuroscience course at my school (which has no physics prerequisites, just math) seem to suggest it's just mathematical/computer modeling of data.

Also I'm deciding which of math, computer science, and physics to double major in. My school has a dual degree program in computer and cognitive science, which is essentially a math and comp sci double major, that is supposed to prepare one for "the science of mental information," so would that be the way to go? I'd hate to give up on physics though, although I am taking an honors mechanics & em sequence this year along with a principles of modern physics, and I have enough room to take some quantum and stat mechanics senior year. Should I do that?

Any advice would be greatly appreciated.

 

[Simfish]

I think the computer and cognitive science double major would be good. You don't have to triple major with physics - you can still take physics courses or self-study them.

Here are two theoretical neuroscientists who have interesting biographies: (you should look at what they did)

http://www.amath.washington.edu/~etsb/
http://www.klab.caltech.edu/~koch/

 

[story645]

Try to sneak in a psych course or bio course on neuroscience just so that you can get a different perspective on the field.

Are neurons actually modeled based on fundamental physics?

Kind of. The models I've seen rely a lot on circuit theory: basically every signal has a small voltage, and a lot of signals travel to any node (neuron) and once the voltage at the neuron reaches the action potential level, the neuron propagates the signal to the next neuron in the chain (which in turn means that neurons are sort of acting as amplifiers). This idea of action potential works the same way in AI neural nets: wait 'til the signal is strong enough, then do some thing. (Usually, action potentials are some kind of probabilistic formula.)

 

[feuxfollets]

Hmm okay. how is computer science used? It seems like most stuff from comp sci would be too abstracted to be of much use in modeling stuff at the neuron level.

@Simfish: it's not actually a cog sci major. I have a choice of majoring in either math, psych, philosophy, linguistics, or biological basis of behavior, along with a comp sci major. And I would almost definitely choose math out of those.

 

[Math Is Hard]

Hmm okay. how is computer science used? It seems like most stuff from comp sci would be too abstracted to be of much use in modeling stuff at the neuron level.

You might be interested in a course on artificial neural networks:
http://en.wikipedia.org/wiki/Artificial_neural_network

 

[Simfish]

Computational biology is inclusive of things at the neuron level. Well, there actually is not a significant difference between theoretical computer science and math. Maybe the more applied parts of CS aren't as useful for theoretical neuro, but theoretical computer science is very useful for things at the neuronal level (for example: pattern classification and neural networks are both studied more in CS departments than math departments)

 

[story645]

Maybe the more applied parts of CS aren't as useful for theoretical neuro,

If you're doing more applied cog-neuro, it's super useful to know some of the applied stuff, 'specially stats. My undergrad psych thesis was almost all signal processing, and if you talk to cog-neuro people you'll consistently hear that they want people with a background in EE or math 'cause of the models. I picked up the signal processing as much from working in a CS lab that does computer vision/image processing as from any of my EE courses. (Which is another thing: if your in CS and there are no signal processing courses, take the image processing/computer vision course-audio, visual, and electric signals are processes in much the same way.)

pattern classification and neural networks are both studied more in CS departments than math departments

Those are both considered applied in CS, unless you're looking at formal verification of the algorithms or their complexities or the like.

how is computer science used?

You'll see this more if you also take a philosophy course with a cognitive bent, but a lot of the really abstract CS stuff-things like formal logic, state machines, turing machines, theories of computation, standard proof methods (induction, contradiction, pigeon hole, etc.)-comes into play in the various theories of cognition.

Even programming concepts can help things gel: my friends and I played around with the idea of modeling the soul as a DB where every life is a different row on the DB and understanding the idea of objects and classes help you understand some of big theories in cognition. (Sorry, can't remember which theories specifically).

 

[cosmoKaty]

If you want to do any research relating to neurology, the lab work will include using or generating computer programs designed to analyse data, for example the electrical signals.

If you enjoy working with computers, getting a degree in computer programing/science will defiantly help you in the work world.

 

[feuxfollets]

Are the abstract math stuff, like real analysis, useful at all or potentially useful?
And I feel like it would easier to self study abstract math stuff after having some experience in them than it would to self study physics.

So I'm kinda thinking of doing physics and computer science now, and just taking some of math/EE stuff as well. I've changed my mind about twenty times in the past two weeks though between those four... how should I go about deciding? Next fall I'll need to start taking major specific tracks so I have to decide by then...