Machine learning without much CS background

[tabis]

Hi,

So so far I've gotten into a few stat PhD programs in the 10-20 range (u.s. news), and in picking a program I'm trying to think of a few research areas that interests me. My background is in pure math and econ but I want to do something applied, other than in econ or finance. Applications to the biological sciences is one area I'm looking at, and I know that a biology background isn't needed for this.

 

On the other hand I'm also interested in machine learning, but I'm not sure whether this is a viable option.

My CS background is only an intro programming course with matlab, a second course with C, and an intro numerical analysis course, nothing else.

 

So here's my question - how hard would it be to succeed in ML with only that much CS background?

It seems that most people who go that route have at least a minor in CS or EE.

Is it actually common for people without much CS to do ML, or would it be more realistic to disregard ML when picking a program?

 

Any thoughts will be greatly appreciated, thanks!

 

Edited March 5, 2013 by tabis

[econ_to_stats]

I'm in a similar boat.  Interested in Machine/Statistical Learning but don't have much of a programming background.  I know R very well and have some statistical computing experience but not much beyond that.  I'm hoping I can just pick up what I need as I need it.  

[sisyphus1]

You can do ML without a CS background. MATLAB should be a minimum though. And a class on algorithms (so you can get familiar with complexity analysis) would be also useful.

 

Having said that, the best ML departments have a lot of CS/EE faculty (Stanford, CMU etc.)

[cyberwulf]

ML is a very, very broad area at this point, spanning fairly abstract statistical theory to quite applied computational methods. With a basic knowledge of programming, I think a smart person could pick up the requisite skills.

[echlori]

I'm from a similar background to some of you, and I'm currently in a short RA stint working on machine learning. So far, it hasn't been too bad. Knowing MATLAB is useful. Going through the relevant Coursera courses will be very helpful, e.g. Andrew Ng's and Geoffery Hinton's courses.

[tabis]

Thanks for the input everyone! I feel a little encouraged now

[tarrman]

ML is more like theoretical CS, so programming knowledge isn't necessary. Many of the professors doing research in it now come from a mathematical background, so I think you'd be plenty prepared.

 

My plan is to do ML as well, but I feel like I'm at a bit of a disadvantage since I don't come from a mathematical background. I'm in a statistical learning class now, and my friend (math major) is able to breeze through the HW with much less effort, while I'm struggling to make sense of all the linear algebra/statistics.