Best preparation for CS PHD programs

[geekman]

Hey guys. So, here is my predicament. I did my undergraduate in Pure Math and just got my masters in Stats. I recently decided a few months ago that I actually want to get into Comp Sci - specifically Machine Learning.

I applied to some PHD programs and so far all I've gotten is rejections. I think reality has set in, and I need more CS experience.

So, what is the best option? Either, I can go for a masters in Computer Science (which I don't want to do unless it is fully funded). Also, I don't know of any fully funded programs that are decent. I can spend a year and take undergrad CS classes and (hopefully) get some research experience and good letters of rec. Or, I can work. I feel like working would be kind of pointless in strengthening my resume.

Any advice is greatly appreciated! Thanks!

[newms]

Your Math and Stats background is really good for Machine Learning, so that is a positive. Your case reminds me of Christopher Scaffidi - he had a background in Physics (I believe he had a MS in Physics) before he decided he wanted a PhD in Computer Science. He did the CS GRE and scored very highly on it and got into Carnegie Mellon and he's now a prof at Oregon State. So you could consider doing the CS GRE since it tests a wide background of undergraduate CS knowledge. That test is pretty hard though and will require hundreds of hours of preparation for someone without a CS background, but doing well on it would show your competence of undergraduate CS topics.

Alternatively, you could consider some mathematically slanted computer science programs. I'm thinking specifically of the University of Alberta, that has a very good machine learning group. They have a Statistical Machine Learning program that is geared towards people with either a Maths background or a CS background, so you can look into that. I'm sure there are other similar programs out there as well. Best of luck!

[Machine_Learner]

Correctly pointed out by newms.

Also, most top Canadian CS schools give funding for MS students

[newms]

I thought of this thread when I was at a recruiting event for a top 20 CS program and some of the admitted PhD students there didn't have a CS bachelor's but had a BS in Maths or Physics. I didn't ask what they did to show their competence in CS background, but it definitely is possible to get in to a good CS program without a CS bachelors.

[victor.s.andrei]

Hey guys. So, here is my predicament. I did my undergraduate in Pure Math and just got my masters in Stats. I recently decided a few months ago that I actually want to get into Comp Sci - specifically Machine Learning.

Kudos on your master's degree.

So, what is the best option? Either, I can go for a masters in Computer Science (which I don't want to do unless it is fully funded). Also, I don't know of any fully funded programs that are decent. I can spend a year and take undergrad CS classes and (hopefully) get some research experience and good letters of rec. Or, I can work. I feel like working would be kind of pointless in strengthening my resume.

Working is not pointless if it's in a research-oriented position (that's at least tangentially related to what you want to do) at a big corporation that's paying for your master's degree from computer science at a mid-range public university.

Go find that employer, and get them to pay for your master's in computer science or at least a few basic non-degree courses (algorithms, operating systems, computer networks).

Then, try again.

[Azazel]

So, what is the best option? Either, I can go for a masters in Computer Science (which I don't want to do unless it is fully funded). Also, I don't know of any fully funded programs that are decent. I can spend a year and take undergrad CS classes and (hopefully) get some research experience and good letters of rec. Or, I can work. I feel like working would be kind of pointless in strengthening my resume.

I would take an another year and do CS research, maybe taking a few classes on the side. I've come to believe that LOR are the most important part of your application. Get good ones, and you are set. Generally speaking, the best way to get useful letters is to do great research.

You might also consider working for a bit, if your job is "researchy" enough and you are working under a current CS PhD. But if your goal is to get into the best grad school possible, I'd be skeptical of this option.

Edited April 20, 2011 by Azazel

[OH YEAH]

Princeton fully funds (and pays a stipend to) all masters students. I think UIUC does something similar?

[qqyyzz]

This is all anecdotes, but from what I've heard many of the people studying machine learning come from math backgrounds and many places will look very specifically at math background for machine learning. Unfortunately, every CS program I have looked at requires (obviously) a bunch of CS courses as admission requirement. I'm not sure how people without a CS major get into these programs; I assume that many of these people majored in something else but have close to the number of CS courses as a major.

Alternatively, you could look for statistics phd programs that do research in machine learning/statistical learning. After all, machine learning is a field that's at the intersection of computer science and statistics and many of the machine learning algorithms are the same as statistics techniques with different terms.

[colincsl]

I'm not at all active on these forums, but I thought I would chime in. I am currently a senior in mechanical engineering and was accepted to several computer science PhD programs for this coming fall.

The reason I think I was accepted is because of excellent recommendations and a plethora of research projects that involve programming. While my background is in mech engineering, I have done of lot of engineering club and research projects where I end up programming some algorithm or putting together some software system. It also helps that my interests (robotics, computer vision) are very interdisciplinary. I actually found during my visits that many students didn't come in with a major in computer science. Several people were computer engineering (essentially half electrical engineering and half CS) or other STEM majors.

The fact that you have a background in statistics and want to go in to an area like machine learning should push your application. It's not like you want to do CS theory where a computer science background is vital. As long as you are confident in your programming abilities and have enough math background then it shouldn't be a big deal.