Positions

The LIONS group at EPFL has several openings for postdoctoral fellows and PhD students for research in machine learning and information processing. Please see our research interests.


For the postdoctoral positions, candidates should have or be close to finishing a PhD degree in electrical engineering, computer science, applied mathematics, or a related field. Candidates should send their CV, a research statement outlining their expertise and interests, any supplemental information, and a list of at least three references with full contact information to the LIONS Lab Administrator:

Gosia Baltaian (gosia.baltaian@epfl.ch)

EPFL STI IEL LIONS
ELD 244 (Bâtiment ELD)
Station 11
CH-1015 Lausanne


For the PhD student positions, we are interested in students with EE, CS, and Mathematics backgrounds. Candidates should directly apply to the EDEE or EDIC doctoral programs and list Prof. Volkan Cevher as a potential host for their PhD studies.

PhD Positions in Machine Learning

Laboratory for Information and Inference Systems (http://lions.epfl.ch/) at Ecole Polytechnique Federale de Lausanne (EPFL) is currently looking for multiple PhD students in Machine Learning.

Informal inquiries should be sent to Gosia Baltaian, gosia.baltaian@epfl.ch.

Topics: 

1. Guaranteed accuracy for machine learning models of materials compound space

The high-throughput screening of large databases of novel materials candidates constitutes a central goal of the recently awarded MARVEL NCCR grant. Given a training set of compounds with pre-calculated quantum mechanical properties, we seek to construct supervised machine learning models that accurately infer the corresponding properties for similar materials with correctness guarantees.

2. Bayesian optimization / Gaussian process bandit optimization

We would like to build an active learning framework based on Bayesian optimization that adaptively queries an unknown function in order to build an explicit approximation or to optimize the function with theoretical guarantees. For this purpose, we would like to unify key combinatorial structures (e.g., submodularity) with smoothness models (e.g., Gaussian processes) for rigorous guarantees.

3.

/* Style Definitions */ table.MsoNormalTable {mso-style-name:”Tableau Normal”; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-parent:””; mso-padding-alt:0cm 5.4pt 0cm 5.4pt; mso-para-margin-top:0cm; mso-para-margin-right:0cm; mso-para-margin-bottom:10.0pt; mso-para-margin-left:0cm; line-height:115%; mso-pagination:widow-orphan; font-size:11.0pt; font-family:”Calibri”,”sans-serif”; mso-ascii-font-family:Calibri; mso-ascii-theme-font:minor-latin; mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin; mso-fareast-language:EN-US;} Discrete optimization with emphasis on submodularityompressive sensing techniques 

/* Style Definitions */ table.MsoNormalTable {mso-style-name:”Tableau Normal”; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-parent:””; mso-padding-alt:0cm 5.4pt 0cm 5.4pt; mso-para-margin:0cm; mso-para-margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:10.0pt; font-family:”Times New Roman”,”serif”;}

/* Style Definitions */ table.MsoNormalTable {mso-style-name:”Tableau Normal”; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-parent:””; mso-padding-alt:0cm 5.4pt 0cm 5.4pt; mso-para-margin-top:0cm; mso-para-margin-right:0cm; mso-para-margin-bottom:10.0pt; mso-para-margin-left:0cm; line-height:115%; mso-pagination:widow-orphan; font-size:11.0pt; font-family:”Calibri”,”sans-serif”; mso-ascii-font-family:Calibri; mso-ascii-theme-font:minor-latin; mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin; mso-fareast-language:EN-US;}

We would like to exploit the underlying combinatorial structures of decision formulations in order to obtain scale up sampling, inference, and decision systems. For this purpose, we would like to leverage submodularity and develop efficient algorithms with provable guarantees.

 4. Scalable convex optimization

We would like to exploit the underlying convex geometry of learning formulations in order to obtain massive speed-ups in learning with convex optimization. The student will work with streaming data models, stochastic approximation, primal-dual smoothing to design new, heuristic-free algorithms with theoretical guarantees.

The LIONS lab provides a fun, collaborative research environment with state-of-the-art facilities at EPFL, one of the leading technical universities worldwide. EPFL is located in Lausanne next to Lake Geneva in a scenic setting with excellent transport connections. The working language at EPFL is English.

Successful applicants need to be highly motivated, excellent students with a solid background in information theory, optimization, computer science, or applied mathematics. Advanced coding skills is a big plus.

Candidates should directly apply to the EDEE or EDIC doctoral programs and list Prof. Volkan Cevher as a potential host for their PhD studies.

The working language at LIONS is English.

Starting date: Continuous

For more details please check: http://phd.epfl.ch/application

Postdoc Positions

The Laboratory for Information and Inference Systems (LIONS) at EPFL is looking for postdoctoral fellows with a strong theory background in machine learning, discrete optimization, information theory, statistics, compressive sensing, or other related areas. Strong coding skills is a big plus.

There are two positions that revolve around the following two topics: 

1) Bayesian optimization, bandits, and reinforcement learning

We seek to develop online algorithms for Bayesian optimization, as well as related problems such as multi-armed bandits, level-set estimation, and reinforcement learning.  The algorithms will be characterized theoretically, and also tested in real-world applications including automated hyperparameter optimization with neural networks and personalized education.

2) Magnetic resonance imaging (MRI) with compressive sensing techniques 

/* Style Definitions */ table.MsoNormalTable {mso-style-name:”Tableau Normal”; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-parent:””; mso-padding-alt:0cm 5.4pt 0cm 5.4pt; mso-para-margin:0cm; mso-para-margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:10.0pt; font-family:”Times New Roman”,”serif”;} We seek to develop techniques in order to optimally subsample the k-space by desigining realistic scan trajectories and waveforms that can be implemented on hardware. Different set-up are to be investigated: multi-coil MRI, dynamic MRI, 2D/3D imaging etc. Improving upon the existing non-linear decoders and speeding up the reconstruction algorithms is another aspect of the project. For this position, a vast experience with MRI hardware and with compressive sensing reconstruction on real data is indispensable.

LIONS provides a stimulating, collaborative and fun research environment with state-of-the-art facilities at EPFL. Personal initiative and independent research tasks related with the candidate’s interests are also encouraged.

The working language at EPFL is English.

Informal enquiries can be sent to Gosia Baltaian, gosia.baltaian@epfl.ch