This article is part of a series of articles on different European master’s programs related to artificial intelligence and machine learning.
Let me preface this writeup by saying that I did not do a master’s degree at EPFL and unfortunately, I have no special insights on the minutiae of a study plan. To avoid wrongly conveying factual information, I will not be discussing specific details of academic programs, and I will focus more on machine learning research and the broader environment around EPFL with an emphasis on aspects of being a student that cannot easily be learned from an internet search.
Masters degrees at EPFL
EPFL is completely a technical university, all academic departments are STEM-y. The degrees are academic and research-oriented (like most degrees in Europe, but different to some shorter and more vocationally-oriented masters degrees that can be found in the US). As I understand it, a master’s degree at EPFL is generally composed of 120 ECTS, with something around 40% of the credits coming from research projects and a master’s thesis, and the remainder coming from coursework, of which about one third will be required courses.
More details on the requirements for master’s degrees can be seen in the program overview.
Machine learning at EPFL is fairly dispersed across departments—many engineering and science departments offering their own variants. There is also a complete list of programs.
My impression is that the bulk of the teaching and research is done in the electrical engineering and computer science departments, but realistically I think that a determined student could be a solid ML education in many of these programs. For example, I recently TAed for a large “Deep Learning” course offered by the EE department, and we had students from about 10 different engineering and science disciplines, math, and computer science. About one-third of the students hadn’t done their bachelor’s degrees at EPFL.
Academics: Research and Coursework
EPFL is generally a well-regarded university. Along with ETH, it benefits from Switzerland’s strong economic situation and high quality of life that tend to attract good faculty and students. Switzerland emphasizes science and technology, and the perception is that although EPFL and ETH are public institutions they have excellent funding and support. My impression is that within the broader CS world, it is best known for producing Scala, though it seems to do okay in ML (FWIW states that it is 14th in the world and second in Europe according to a methodology described there).
EPFL does not have any “household names” in ML, but plausibly more importantly for a master’s student, we have many younger, enthusiastic faculty who are genuinely quite nice (my impression is that Swiss culture resists big egos and people who are too self-important). More generally, the university is quite young, and IMHO this can be felt in a positive way. In short, the academic atmosphere is quite nice!
Here is a list of faculty that the computer science department considers being doing work in different areas.
Research is a definite focus and access to faculty is generally fairly good. Masters degree students will do projects in collaboration with research groups, this can oftentimes mean having a desk physically colocated with the lab, and participating with the activities of the lab. This is a natural route to publishing research and progressing on to a Ph.D.
It is common to see postings around the university for masters degree projects in collaboration with private companies, very often start-ups spun out of the EPFL. And there are other natural venues for communicating fruitfully with practitioners, such as the “Swiss Data Science Center” and the “Applied ML days” conference. CERN (particle accelerator), Idiap (semi-private AI research institute), CHUV (world-class research hospital), Armasuisse (a government agency that works with researching military technologies), and a host of international organizations (for example, much of the policy discussion about autonomous weapons focuses around the UN in Geneva) are other important sources of demand for machine learning talent in the area.
Unfortunately, I cannot comment on the minutiae of specific courses. Broadly, my qualitative impression is that they are taught to the highest international level. EPFL is active on MOOCs, so in principle, anyone can get a sense of the course. Another view on EPFL might be seen from the quality of universities it officially partners with.
The EPFL itself is not outstandingly beautiful: the architecture is fairly brutal and the facilities are not the newest. It’s not ugly, though, either, it’s clean and well-run and generally feels safe and comfortable. Campus life is not outstanding. There are some activities, but it is much less emphasized than, for example, at US research universities or Oxbridge. Much of the student body is from the surrounding area (it is feasible to commute from essentially all of French-speaking Switzerland), and public transportation is excellent such that the incentive for those without strong preferences to live within walking distance of campus is low. One consequence is that the campus can be quite empty in the evenings and weekends.
Basically, the campus is located in a suburb of nearby Lausanne. It’s a nice area, but largely housing, without a lot of entertainment or dining. One redeeming feature is beautiful Lake Geneva, which is about a 10-minute walk from campus. There is a large sports center shared with the neighboring University of Lausanne. Downtown Lausanne, which has all of the amenities of a world-class city, is about a 20-minute local train ride. Geneva Airport, which has excellent connections, is about 70 minutes by train. There are excellent outdoor activities within easy reach and, for example, one can see advertisements for hikes or skiing organized by student groups.
There are many international students, and EPFL participates aggressively in exchange programs with other universities. Unsurprisingly, the student body skews European, but the Geneva area is famously cosmopolitan, so cultural fit should be less of a problem at EPFL than probably most other places in continental Europe.
Anecdotally, it seems common to do an internship as part of one’s studies. As far as I’m aware, there’s no Lausanne branch of Google or Amazon or any other mega tech companies, but there is a quite active startup scene in the area. Collocated with the EPFL campus is a large Silicon Valley-like innovation park, and there are many English-language groups, e.g. this, around entrepreneurship in ML. Geneva and Switzerland more generally have thriving finance, pharmaceutical, and technology sectors.
There is an active AI safety reading group: five to ten people who get together each week to discuss recent research pertinent to AI safety.
EPF Lausanne, ML related MSc programs
By Kyle Matoba
This article is part of a series of articles on different European master’s programs related to artificial intelligence and machine learning.
Let me preface this writeup by saying that I did not do a master’s degree at EPFL and unfortunately, I have no special insights on the minutiae of a study plan. To avoid wrongly conveying factual information, I will not be discussing specific details of academic programs, and I will focus more on machine learning research and the broader environment around EPFL with an emphasis on aspects of being a student that cannot easily be learned from an internet search.
Masters degrees at EPFL
EPFL is completely a technical university, all academic departments are STEM-y. The degrees are academic and research-oriented (like most degrees in Europe, but different to some shorter and more vocationally-oriented masters degrees that can be found in the US). As I understand it, a master’s degree at EPFL is generally composed of 120 ECTS, with something around 40% of the credits coming from research projects and a master’s thesis, and the remainder coming from coursework, of which about one third will be required courses.
More details on the requirements for master’s degrees can be seen in the program overview.
Machine learning at EPFL is fairly dispersed across departments—many engineering and science departments offering their own variants. There is also a complete list of programs.
My impression is that the bulk of the teaching and research is done in the electrical engineering and computer science departments, but realistically I think that a determined student could be a solid ML education in many of these programs. For example, I recently TAed for a large “Deep Learning” course offered by the EE department, and we had students from about 10 different engineering and science disciplines, math, and computer science. About one-third of the students hadn’t done their bachelor’s degrees at EPFL.
Academics: Research and Coursework
EPFL does not have any “household names” in ML, but plausibly more importantly for a master’s student, we have many younger, enthusiastic faculty who are genuinely quite nice (my impression is that Swiss culture resists big egos and people who are too self-important). More generally, the university is quite young, and IMHO this can be felt in a positive way. In short, the academic atmosphere is quite nice!
Here is a list of faculty that the computer science department considers being doing work in different areas.
There is also this cross-disciplinary page meant to emphasize ML across the university.
Research is a definite focus and access to faculty is generally fairly good. Masters degree students will do projects in collaboration with research groups, this can oftentimes mean having a desk physically colocated with the lab, and participating with the activities of the lab. This is a natural route to publishing research and progressing on to a Ph.D.
It is common to see postings around the university for masters degree projects in collaboration with private companies, very often start-ups spun out of the EPFL. And there are other natural venues for communicating fruitfully with practitioners, such as the “Swiss Data Science Center” and the “Applied ML days” conference. CERN (particle accelerator), Idiap (semi-private AI research institute), CHUV (world-class research hospital), Armasuisse (a government agency that works with researching military technologies), and a host of international organizations (for example, much of the policy discussion about autonomous weapons focuses around the UN in Geneva) are other important sources of demand for machine learning talent in the area.
Unfortunately, I cannot comment on the minutiae of specific courses. Broadly, my qualitative impression is that they are taught to the highest international level. EPFL is active on MOOCs, so in principle, anyone can get a sense of the course. Another view on EPFL might be seen from the quality of universities it officially partners with.
Or, more directly, you might cross-check publicly available course materials with the degree requirements of a program you are interested in (see e.g. the 2019 ML course, the Deep Learning course and list of courses).
Student life and logistics
The EPFL itself is not outstandingly beautiful: the architecture is fairly brutal and the facilities are not the newest. It’s not ugly, though, either, it’s clean and well-run and generally feels safe and comfortable. Campus life is not outstanding. There are some activities, but it is much less emphasized than, for example, at US research universities or Oxbridge. Much of the student body is from the surrounding area (it is feasible to commute from essentially all of French-speaking Switzerland), and public transportation is excellent such that the incentive for those without strong preferences to live within walking distance of campus is low. One consequence is that the campus can be quite empty in the evenings and weekends.
Basically, the campus is located in a suburb of nearby Lausanne. It’s a nice area, but largely housing, without a lot of entertainment or dining. One redeeming feature is beautiful Lake Geneva, which is about a 10-minute walk from campus. There is a large sports center shared with the neighboring University of Lausanne. Downtown Lausanne, which has all of the amenities of a world-class city, is about a 20-minute local train ride. Geneva Airport, which has excellent connections, is about 70 minutes by train. There are excellent outdoor activities within easy reach and, for example, one can see advertisements for hikes or skiing organized by student groups.
There are many international students, and EPFL participates aggressively in exchange programs with other universities. Unsurprisingly, the student body skews European, but the Geneva area is famously cosmopolitan, so cultural fit should be less of a problem at EPFL than probably most other places in continental Europe.
Anecdotally, it seems common to do an internship as part of one’s studies. As far as I’m aware, there’s no Lausanne branch of Google or Amazon or any other mega tech companies, but there is a quite active startup scene in the area. Collocated with the EPFL campus is a large Silicon Valley-like innovation park, and there are many English-language groups, e.g. this, around entrepreneurship in ML. Geneva and Switzerland more generally have thriving finance, pharmaceutical, and technology sectors.
There is an active AI safety reading group: five to ten people who get together each week to discuss recent research pertinent to AI safety.