Saarland University, Machine Learning Group, Fak. MI - Mathematik und Informatik, Campus E1 1, 66123 Saarbrücken, Germany     

Machine Learning Group
Department of Mathematics and Computer Science - Saarland University

TEACHING

SEMINAR ON "ADVANCED TOPICS IN MACHINE LEARNING"

Seminar in summer semester 2017

DESCRIPTION

The seminar concentrates on reviewing recent developments in machine learning and related topics from the literature such as optimization and computer vision. The seminar focuses on master/phd students with background in computer science. Almost all the papers we are going to review in this seminar were accepted papers at top conferences in machine learning like NIPS, ICML, COLT.

SEMINAR FORMAT AND REQUIREMENTS

The seminar will run in reading group format. There is only one session per week where the student presents his/her corresponding paper. The presentation of each paper should take only 30 minutes and should also include a review of the relevant background that are needed to understand the presented topics. Every session will also have upto three students (assigned by us) who will lead an informal discussion of the paper. At the end of the seminar, all participants are required to submit a short summary (approx. 4 pages) of their assigned papers.

PREREQUISITES

Familiarity with the basics of machine learning, probability theory/statistics and optimization is a plus but not strictly required.

PAPERS

Date Speaker Paper Tutor Opponents Links
22.05. Y. Fan Generative Adversarial Nets.
NIPS 2014.
QN V. Lazova
M. Augustin
29.05. V. Lazova Regularization of Neural Networks using DropConnect. ICML 2013. QN M. Andriushchenko
12.06. M. Augustin Globally Optimal Training of Generalized Polynomial Neural Networks with Nonlinear Spectral Methods. NIPS 2016. AG V. Lazova
P. Kolev
19.06. M. Andriushchenko Explaining and harnessing
adversarial examples. ICLR 2015
MH Y. Fan
D. Mehta
26.06. D. Mehta Coordinate-wise Power Method.
NIPS 2016.
AG Y. Fan
M. Augustin
03.07. S. Mahajan Why should I trust you? Explaining
the predictions of any classifier.
KDD 2016.
MH D. Mehta
P. Kolev
17.07. P. Kolev Gradient Descent only Converges
to Minimizers. COLT 2016
QN M. Andriushchenko

TIME AND LOCATION

First meeting: 05.05.2017, 10:00, E1 1, SR 206

Seminar: Monday, 14-16, E1 1, SR 206

LECTURER

Prof. Dr. Matthias Hein

Quynh Nguyen

Antoine Gautier

Office Hours:

 

NEWS

Registration: Please register for the seminar on HISPOS. Deadline is Monday, June 5, 2017.

Write-up: The write-up (PDF format) must be sent to your tutor via email until 17.07.2017. It should summarise your talk as well as the main points of the paper (max. 4 pages).  Do not forget to hand in your writeup: Participants who do not submit a writeup cannot obtain the certificate for the seminar.