How useful is control theory in computational neurosciences?

[fatpotato]

Hello,

I have the opportunity of choosing an elective course on control theory before embarking for my master's program in computational neurosciences and I am weighing the pros and cons. I am looking for advice from PF members who specialized in this field.

Since I have chosen enough courses to validate my program, adding this course to the pile would slightly overload my schedule and require both taking an exam and attending practical labs with lab reports to write.

I find control theory extremely stimulating, but in this particular case, I would only consider taking it if there is a good return on investment later, so here is my question: how useful is control theory in computational neurosciences and is it worth taking this course?

Thank you.

 

[DaveE]

First, I know little to nothing about computational neurosciences, except my vague impression that neural stuff tends to be very non-linear. OTOH, control theory, which I did study is initially all about linear systems (linear algebra, complex analysis, and such). Non-linear control theory is pretty advanced in the catechism, and frankly, probably not that useful. It's really interesting, but hard to apply in practice.

However, my experience doesn't really include the "computational" aspect, it was mostly math. Honestly, I don't know what "computational" control theory is. When I was studying 30 years ago, this was an oxymoron. The computational part was to replicate in a machine the mathematical solutions the controls guy did.

I would suggest you look into some of the online lectures from places like MIT to see if they are covering material that you would use. Don't listen to all of the lectures, just skim through as necessary to see if you care about what they are covering. You could also look at some modern controls textbooks, like the TOC from Amazon, to see what is typically covered.

 

[fatpotato]

Thank you very much for your advice.

I also have the impression that the brain tends to be quite non-linear, but the following reference surprised me. It is funny that you mention MIT, because they apparently published a book called "Neural Control Engineering, the emerging intersection between control theory and neuroscience".

This alone would answer my question, but I suspect that the topics covered in this book go way above what I would go through during a master's program.

My problem boils down to: "I don't know enough about neurosciences to correctly assess if knowing control theory would be a decisive advantage".

 

[Joshy]

I don't know anything about neuroscience myself, but my undergraduate required me to do a depth or concentration within the major, and I did mine on controls.

My controls classes were a lot of block diagrams we started off quickly describing a set of these blocks into a transfer function that would represent a linear system then the professor might ask us to do something like describe a characteristic such as damping factor or response to a step, or if it was stable or not. We learned how to draw and interpret a few graphs like Nyquist contour, Nichol's chart, or root locus. The later classes covered things like state space systems lots of matrices... system looked partition between a hardware side and a software side or that's how I interpreted.

The state space stuff in the more advanced coursework I never see it anymore even though I thought that was more interesting material. The stuff in the first class like stability and characterizing I see all the time even outside of controls, and I would imagine (just a speculation here) that a lot of other areas of study might use similar techniques to describe an overall system. The class started off with 2nd order systems, but later into that class we might look for like a dominant pole pair for higher order systems and it might be close enough. Now I see it with all sorts of things that have some type of feedback or sensing.

So I think the class wouldn't be a total waste, but if I could back in time and do things differently myself... I had this very impractical and silly way of looking at my education just wanting to take every class and learn everything I can... It's okay to learn, but I spread myself too thin and wasn't acing my classes on the way out. For me... I probably should have just taken the classes I needed then while it might not be as good as learning in from a class from the professor if it's just extra material "nice to haves" that I wanted to learn for fun then I think it's okay not to do it the most rigorous way. It sounds like it's very hard for you to make the connection between your goals and this class so I probably would put it in the nice to haves category for now just focus on doing a good job with what you need then if there's ever any time or an opening maybe even during you graduate studies you can add that in there later.

 

[fatpotato]

Thank you for your feedback.

I had this very impractical and silly way of looking at my education just wanting to take every class and learn everything I can... It's okay to learn, but I spread myself too thin and wasn't acing my classes on the way out.

Same for me. I do not know if we are similar, but I would personally take all the electives if I had enough money to stay in school. Alas, I also tend to spread myself thin and I would indeed not be able to perform well everywhere.

The later classes covered things like state space systems lots of matrices... system looked partition between a hardware side and a software side or that's how I interpreted.

The state space stuff in the more advanced coursework I never see it anymore even though I thought that was more interesting material.

I agree, state space methods are extremely interesting, and I would be delighted to discover if they can be applied to neurosciences. Basically, I am looking for any excuse to use them.

Ok, I am now strongly considering keeping this course on the side to self-study, but still, if any PF member with a background in neuro. can share their experience, please do.

 

[berkeman]

Thank you for your feedback.

I see what you did there.