Winter School on Deep Learning

21st to 25th January

Paolo Favaro

Prof Paolo Favaro, University of Bern

Dr Francois Fleuret, IDIAP, Martigny

Prof. Marcus Liwicki, Luleå University of Technology, Sweden and UNIFR

Dr Moustapha Cisse, Google, Dr Soumith Chintala, Facebook, Dr Jan Koutnik, NNAISENSE M. Klaus Greff, IDSIA, Switzerland Prof. Alyosha Efros, University of Berkeley, Prof Rene Vidal, John Hopkins University, Dr Francois Fleuret, IDIAP, Martigny, Prof. Paolo Favaro, UNIBE

Overview: Deep learning (DL) is today an extremely active research field, whose developments in the last 12 years have drastically increased the performance of data analysis methods and promise to bring further high-impact solutions to data-rich domains. Moreover, because many fields of research are becoming more and more data-driven and DL methods are data-agnostic, DL technologies see widespread application to academic research and the industry beyond computer science. As the scientific community is heavily investing efforts in DL, the field is progressing at an unprecedented pace and it is therefore imperative to have events (such as this Winter School), where the latest and most important results are presented in a structured format by leading researchers in the field. Goal of the Winter School: In this CUSO Winter School we will bring together international experts in deep learning to share their expertise and provide a first-hand account of the fascinating developments and opportunities in this area. The goal of the school is twofold. First, the school will offer an introduction and overview of the research area for a broader audience, illustrating the relevance of the topic in scientific and commercial applications. In addition, for more advanced students it will foster interaction with renowned experts in the field, supporting them to become active and independent researchers on their own. We believe this winter school will address a broad audience of doctoral students and postdoctoral researchers at the CUSO universities, which are home to a number of research groups in related areas. Organizational Details: Date: Monday through Friday, 21.-25. January 2019, Location: Hotel Kreuz, Lenk, Schedule: two lecture sessions from 8:30 to 12:00 and 16:30 to 20:00, afternoon off for social activities; starting Monday afternoon through Friday morning for a total of 8 session.

More information is available at www.winterschool.inf.unibe.ch

Monday 21st 

12.00 - 12.30 Arrival and registration at the hotel

12.30 - 13.00 Introduction by the organizers (Paolo Favaro and Francois Fleuret)

Welcome to the school and presentation of the schedule and activities

13.00 - 14.30 Theory of Deep Learning - Part I (Rene Vidal)

14.30 - 15.00 Coffee break

15.00 - 16.30 Theory of Deep Learning - Part II (Rene Vidal)

17.00 - 20.00 Attendees poster session

20.15 Dinner

Tuesday 22nd

8.30 - 10.00 Background on DL – Part I (Francois Fleuret)

10.00 - 10.30 Coffee breaks

10.30 - 12.00 Background on DL – Part II (Francois Fleuret)

12.15 lunch

13.00 - 16.30 Free time for social activities

16.30 - 18.00 Applications – Part I (Moustapha Cisse)

18.00 - 18.30 Coffee breaks

18.30 - 20.00 Applications – Part I (Moustapha Cisse)

20.15 dinner

Wednesday 23rd

8.30 - 10.00 Unsupervised learning – Part I (Alyosha Efros)

10.00 - 10.30 Coffee breaks

10.30 - 12.00 Unsupervised learning – Part II (Alyosha Efros)

12.15 lunch

13.00 - 16.30 Free time for social activities

16.30 - 18.00 Reinforcement Learning – Part I (Jan Koutnik)

18.00 - 18.30 Coffee breaks

18.30 - 20.00 Reinforcement Learning – Part II (Jan Koutnik)

20.15 dinner

Thursday 24th

8.30 - 10.00 Generative Learning - Part I (Soumith Chintala)

10.00 - 10.30 Coffee breaks

10.30 - 12.00 Generative Learning - Part II (Soumith Chintala)

12.15 lunch

13.00 - 16.30 Free time for social activities

16.30 - 18.00 RNNs (Klaus Greff)

18.00 - 18.30 Coffee breaks

18.30 - 20.00 The binding problem (Klaus Greff)

20.15 dinner

Friday 25th

8.30 - 10.00 Project presentations – Part I (Paolo Favaro)

10.00 - 10.30 Coffee breaks

10.30 - 12.00 Project presentations – Part II (Paolo Favaro)

12.15 Closure and departure

Lenk

The Winter School will take place at Hotel Kreuz in Lenk. Participants do not need to make their own reservation here. The Winter School automatically takes care of this.

Participants from the CUSO Doctoral Program in Computer Science are asked only to contribute a modest registration fee of Fr. 160.-. Accommodation costs for 4 nights in a double room and all meals are fully sponsored by the CUSO Doctoral Program. In addition, travel costs will be refunded for second class SBB tickets. No reimbursement will be made without the original tickets. You will get instructions during the school on how to get reimbursed.

Participants affiliated with CUSO universities should register directly here in the CUSO webpage. After registering for the winter school, wait for the confirmation of your registration. Only then make the payment for the participation fee to the CUSO account. You will receive bank account details during the registration process.

If you are interested to register for the school but you are not affiliated with a CUSO university, please contact the organizers.

40

For questions please contact Prof. Paolo Favaro, favaro(at)inf.unibe(dot)ch, Prof. Marcus Liwicki marcus.liwicki(at)ltu(dot)se or François Fleuret francois.fleuret(at)idiap(dot)ch.