Postdoc position on structured prediction at MILA, University of Montreal

Applications are invited for a 2-year postdoctoral position in the
Department of CS and OR of University of Montreal, working with Simon
Lacoste-Julien on structured prediction, optimization and/or the theory of deep learning.
** I will be at NIPS in Long Beach from Dec 2nd to Dec 9th, so do not
hesitate to catch me to chat about the position (but please email me a CV first). **
The research project is flexible, but related areas of interest include:
- theoretical analysis of structured prediction approaches (old and
new, including but not limited to deep learning ones)
- optimization algorithms, continuous or with combinatorial objects
- broader theory of deep learning approaches
The candidate will be a member of the Montreal Institute for Learning
Algorithms (MILA - http://www.mila.umontreal.ca/ ), and collaborations
there are encouraged. Close collaborations with Andrea Lodi who leads the
Canada Excellence Research Chair in Data Science for Real-Time
Decision-Making are also possible (
http://cerc-datascience.polymtl.ca/ ).
The MILA offers a world-class research environment, with close collaborations with multiple universities (McGill University, Polytechnique, HEC) as well as private research labs (Google Brain, Deepmind, Microsoft Research, FAIR, Element AI, etc. which all have opened an office in Montreal in the last year).
Candidates mush have a PhD in machine learning or related field such
as statistics or optimization.
Starting date is flexible; it could be as early as January 2018 but
should be earlier than October 2018.
Interested candidates should apply online on the MILA admission website:
https://mila.quebec/en/admission/
(Choose the post-doc section, click on "MILA application form" and follow the instructions; choose "Simon Lacoste-Julien" as the Faculty Adviser. After applying, send me an email (slacoste at iro.umontreal.ca) with [POSTDOC] starting the subject header to warn me that you applied for this position.)
To get a better sense of relevant current research interests, have a look at these papers:
On Structured Prediction Theory with Calibrated Convex Surrogate Losses, A. Osokin, F. Bach and S. Lacoste-Julien, NIPS 2017 (oral on Wed Dec 6th 10:20am-AT-Hall A)
https://arxiv.org/abs/1703.02403
SEARNN: Training RNNs with Global-Local Losses, R. Leblond*, J.-B. Alayrac*, A. Osokin and S. Lacoste-Julien, arXiv:1706.04499, [cs.LG], 2017
https://arxiv.org/abs/1706.04499
Parametric Adversarial Divergences are Good Task Losses for Generative Modeling, G. Huang, H. Berard, A. Touati, G. Gidel, P. Vincent and S. Lacoste-Julien, arXiv:1708.02511 [cs.LG], 2017
https://arxiv.org/abs/1708.02511
A Closer Look at Memorization in Deep Networks, D. Arpit*, S. Jastrzebski*, N. Ballas*, D. Krueger*, E. Bengio, M. S. Kanwal, T. Maharaj, A. Fischer, A. Courville, Y. Bengio and S. Lacoste-Julien, ICML 2017
https://arxiv.org/abs/1706.05394