Optimal Transport for Machine Learning: Theory and Applications


  • Davi Sales Barreira - candidato ao título de mestre


25/03/2021 - 14:00


Via Zoom https://ide-fgv-br.zoom.us/j/99637887773?pwd=Q2M3anFHZjRaeGc5V2k5RmhMZTZtUT09 ID: 996 3788 7773 Senha: 994271


In recent years, advances in Optimal Transport have led to a surge of applications in fields such as Economics, Quantitative Finance and Signal Processing, among others. One area in which it has been found particularly successful is Machine Learning. The development of computationally efficient methods for solving OT problems opened doors for creating machine learning algorithms using concepts from Optimal Transport. These new algorithms encompass many different sub-areas such as Transfer Learning, Clustering, Dimensionality Reduction, Generative Models, just to name some.

This work provides an overview of the different ways in which Optimal Transport has been used in Machine Learning, thus helping Machine Learning researchers to better understand the impact of optimal transport in the field and how to use it. This thesis first introduces the main theoretical and computational aspects of optimal transport theory in an accessible way to machine learning researchers, followed by a semi-systematic literature review focusing on the main uses of optimal transport in machine learning.

Texto informado pelo autor.

Thesis Committee: 

  • Eduardo Fonseca Mendes (orientador) - FGV EMAp
  • Alberto Paccanaro - FGV EMAp
  • Roberto Imbuzeiro - IMPA