CMU CS PhD vs UM Stats PhD

[throwaway9999]

Hi everyone,

 

I have been agonizing over this decision for weeks. I applied to a combination of both CS and PhD programs this season, and got a few offers. Out of my CS admits, CMU CS is my best option and out of my Stats admits, Michigan is currently my best choice. In the beginning, I am pretty much set to go to CMU. However, the more I knew about the Michigan Stats program, the more I think it is a better fit for me. Yet I can not really let CMU go. I am heading to Pittsburgh this Thursday and will go to Ann Arbor on the mid of March for visitation so hopefully I can get a better feelings by then. But for now I would just like to get it off my chest and hear from your (hopefully impartial) opinion.

 

Background: CS senior in the Foundation & Theory track who took at lot of math classes (Cal I, II, II, IV + Linear Algebra), ODE, Abstract Algebra I, Real Analysis I + II, Probability Theory. I am interested in Machine Learning, Bayesian Machine Learning and Bayesian Nonparametric Model.

 

CMU: PhD Computer Science - Theory (not ML or Joint Stat-ML but my POI is an active researcher in Theoretical ML)

(+) Best CS school in the world (I have wanted to go to CMU CS grad school since my 1st year in college)

(+) Probably have the best ML faculty in the US which I can hopefully collaborate

(+) Copious numbers of interesting CS courses

(+) Very good placement record: Have placed one to Stanford and one to Berkeley and one to UChicago for tenured track positions.

(/) Many of my friends go there so I know the place quiet well

(/) No Qual Exam -> could be good or bad depending on your own opinions

(-) Heavily CS oriented curriculum. Please let me explain why it is a negative point. I am admitted into Computer Science PhD program (not ML or Stat-ML) so I have to take breadth courses from 6 main area of CS including Software Systems, Computer Architecture and Programming Language so I will not have enough time to take courses in Probability & Statistical & Measure Theory which are very crucial to ML research.

(-) Top 30 Math department - again, please let me explain why math department has anything to do here. I like mathematics a lot so I would like to get an additional MS in Pure Mathematics during my PhD

 

Michigan: PhD in Statistics

(+) Very good program reputation: most of my professors think very highly of Michigan in general and especially of their Stats program in specific.

(+) Math-Stats curriculum oriented: I only need to take courses in Probability & Statistics & Measure theories so I guess I will have solid foundation to do research in ML.

(+) Perfect Research Fit: My POI at Michigan does exactly what I want to do for my PhD.

(+) Comparable placement record: Surprised to me, Michigan Stat has a comparable placement record to CMU. Even only looking at the CMU ML department, not including one to Stanford and one to Berkeley Assistant Professors, they are similar: 

https://www.ml.cmu.edu/people/alumni-phd.html

https://lsa.umich.edu/stats/alumni-friends/statistics-phd-alumni.html

(+) Have Qual Exam: I like the fact that there are two Qual Exam (Theory + Data Analysis) which helps me to master material better

(+) Top 10 Math department: Since my Stat curriculum is very close to Analysis, it is easy for me to take extra 6 courses in Analysis, Topology and Numerical Method to get an extra MS degree in Math from a very good Math department.

(-) Overall prestige: I know it is petty but I went to college in the North East and not many people know that Michigan is a good school.

(-) Research output: I am not sure if it is reliable but according to this website (http://csrankings.org/) which ranks institutions by number of papers in top tier journals and conferences. For AI field, CMU research output is almost 3 times Michigan (59 vs 20), so I could potentially miss out more interesting projects at CMU ?

 

So am I right to go to Michigan where I think it best-fits me ? Or Am I crazy to turn down CMU offer?

Edited March 1, 2017 by throwaway9999

[abstract_art]

I think since you seem to have a good idea of what you want to do, the overall placements of the program aren't important so much as the placements of the professors you're interested in. Hypothetically, if someone from UM stats was placed at like Stanford or Berkeley but their research area/advisor were in something like spatial statistics, that shouldn't really affect your decision. You seem to have POI at both schools, so look at where their students have published/ been placed.

[throwaway9999]

10 hours ago, marmle said:

I think since you seem to have a good idea of what you want to do, the overall placements of the program aren't important so much as the placements of the professors you're interested in. Hypothetically, if someone from UM stats was placed at like Stanford or Berkeley but their research area/advisor were in something like spatial statistics, that shouldn't really affect your decision. You seem to have POI at both schools, so look at where their students have published/ been placed.

Thanks for the advice ! 

I have the option to be co-advised so while I lam locked down one with one POI, I can still search for an additional advisor. It is quite easy to do so at Michigan.