EPFL vs UIC

[fukuyama]

I am a student completing my M.S. at the University of Minnesota and have had an research assistantship there for nearly three years. My background is in data mining of time series data, and geosocial/social network analysis. I have standing offers from University of Illinois-Chicago under a professor (some weak anonymization for the sake of search terms) who works in modeling of dynamic networks, esp. non-human animal communities and EPFL in Switzerland under a professor there who has recently published a mobile phone dataset collected longitudinally in Switzerland. The former has a background in theory, and is very method-focused, while the latter has a background in computer vision and more core machine learning and is much more application-focused (i.e. "computational social science" they like to call it).

I am not giving much weight to rankings per-se, first of all many rankings are based on journal publications, while computer science is by far driven by conference publications. But as someone with a background in data mining, myself and my peers are much more familiar with the work at UIC. I was drawn to EPFL because of this mobile data project in Lausanne. The researcher at EPFL is still well-cited, is invited for keynotes and these kinds of things in ubicomp, and collaborates with people like Pentland at MIT.

I also prefer EPFL in most aspects: pay, lab facilities, environment, weather. On other aspects, both places seem strong: I really like the working relationship both faculty have with their students. (As someone from Minneapolis) I've always had a strong prejudice against Chicago, and looking for apartments there is very disappointing, personally.

My primary question is: What should be my expectations that my advisor and university should provide, and what should be strictly on me? I mean that, if I have reservations because the faculty at EPFL doesn't publish in conferences I'm used to, should it be my responsibility to strive for these conferences? Second, I would prefer my PhD be more methods-oriented because these are the types of works I most admire in data mining and machine learning. Is it more difficult to pursue this with someone who does not have a strong background in theory? I mean that, both these faculty are mature computer scientists; I feel they both would be able to oversee a wide range related to machine learning.

Thank you