Conference on robustness and privacy

7 students

Modern, huge databases are naturally exposed to corruptions. These may be due to hardware issues such as servers crashing down, databases can be deteriorated during storage, when information is compressed or when messages are exchanged. Another problem is human corruption, that occur involuntarily during their manipulations, yielding wrong  labeling or gross errors and voluntarily  as for fake news. Some corruptions are more subtle as when experiments are slightly perturbed by some cognitive bias. As anomalies cannot be ignored in practice, it is important to study statistical performance of machine learning algorithms on databases possibly containing outliers and/or with heavy-tailed.

Robustness has been studied a lot in statistics since the seminal work of Huber. Many algorithms have also been designed in Machine Learning. However, this subject has witnessed an important renew  during the last 10 years both in the statistical and computer science communities. It involves statistics, optimization, probability and machine learning as mathematical domains. 

In the meantime, privacy has received a lot of attention in the Computer science community because it is a central issue for the security of many sensitive data in finance, economics, administration and, more basically, for keeping customers’ trust in trading. It seems however possible to randomize data in order to ensure a controlled amount of privacy of the individuals while still being able to learn some patterns out of it: this is the cornerstone of privacy. Only very recently, statisticians started to analyze privacy mechanisms. In particular, they discovered several interesting common features between robustness issues and privacy. That is the main motivation on our side to organize a joint conference simultaneously on the two subjects. 

We believe that interesting interaction between people working on robustness and privacy may result in interesting collaborations. These fields have become mature enough in order to organize a three day workshop on this subject. 

 

Organization commitee: Cristina Butucea, Victor-Emmanuel Brunel, Nicolas Chopin, Arnak Dalalyan, Guillaume Lecué, Matthieu Lerasle, Vianney Perchet and Alexandre Tsybakov.

 

Support: The conference is suported by the mathematical institute of CNRS and the Médiamétrie Chair.

For more informations, check homepage of Guillaume Lecué: Conference on robustness and privacy (lecueguillaume.github.io)

  • Table of contents

Instructors

Olivier Catoni

Director of Research CREST, Statistics Laboratory UMR 9194 of CNRS.

Sam Hopkins

Assistant Professor in the Department of Electrical Engineering and Computer Science, MIT.

Pavlo Mozharovskyi

Associate Professor at Télécom Paris in the Team Signal, Statistique et Apprentissage (S2A)  of the Information Processing and Communication Laboratory (LTCI).

Thomas Berrett

Assistant Professor at the Department of Statistics, University of Warwick.

Marco Avella Medina

Assistant Professor in the Department of Statistics at Columbia University.

John C. Duchi

Associate Professor of Statistics and Electrical Engineering at Stanford University

Weijie Su

Assistant Professor of Statistics and Data Science at The Wharton School of the University of Pennsylvania.

Gautam Kamath

Assistant Professor at the University of Waterloo’s Cheriton School of Computer Science, and a faculty affiliate at the Vector Institute.

Clément Canonne

Lecturer in the School of Computer Science of the University of Sydney.

Yeshwanth Cherapanamjeri

Ph.D candidate in Computer Science at UC Berkeley advised by Peter Bartlett.

Jules Depersin

PhD Student in Statistics at the Center for research in economics and statistics (CREST), Paris.

Chao Gao

Assistant professor in statistics at University of Chicago.

Free