Optimal Transport and Analysis for Machine Learning (20w5126)


(University of British Columbia)

Marco Cuturi (Université Paris-Saclay)

Hyung Ju Hwang (Pohang University of Science and Technology)

(Carnegie Mellon University)


The Banff International Research Station will host the "Optimal Transport and Analysis for Machine Learning" workshop in Banff from May 10 to May 15, 2020.

Dealing with large sets of often high-dimensional data, and extracting the information they contain, is an important practical problem that is central to modern machine learning. The theory of optimal transportation (OT) provides a robust way of approaching important aspects of such problems by giving a natural way of measuring differences between datasets. It has had seen remarkable theoretical advances over the recent years, especially finding many surprising connections to analysis and geometry, as well as in understanding the behavior of the optimal transport solutions. These have led a large range of applications, e.g. to economics, image processing, and statistics, although lack of fast and stable numerical algorithms for computing OT correspondences has been a bottleneck for a long time. However, over the last five years there have been breakthroughs, producing new algorithms which have enabled OT to be effectively applicable in data-intensive fields such as machine learning.

Optimal transport has now become an integral part of models for important tasks in machine learning, including deep learning and domain transfer. Recently, analytical approaches have shown to provide important information about the behavior of the functions that model key tasks in machine learning, including their asymptotics and stability, as well as their numerical optimization. The aim of the workshop is to bring together experts, as well as young researchers, in optimal transportation, applied analysis, scientific computing, and machine learning, to consider some of the interesting and deep research questions that would benefit from joint expertise.

The Banff International Research Station for Mathematical Innovation and Discovery (BIRS) is a collaborative Canada-US-Mexico venture that provides an environment for creative interaction as well as the exchange of ideas, knowledge, and methods within the Mathematical Sciences, with related disciplines and with industry. The research station is located at The Banff Centre in Alberta and is supported by Canada's Natural Science and Engineering Research Council (NSERC), the U.S. National Science Foundation (NSF), Alberta's Advanced Education and Technology, and Mexico's Consejo Nacional de Ciencia y Tecnología (CONACYT).