Statistical Physics of Assignment and Optimal Transport: some applications
Assignment and Optimal Transport problems are important tools in many diverse fields such as operational research, signal processing, computer vision, machine learning, etc. We present a study of the finite temperature assignment and optimal transport problems. We show how the statistical mechanics approach provides very fast algorithms to solve these problems and how it outperforms existing algorithms. We show some applications to pattern recognition, and conclude with a discussion of the Wasserstein-Gromov metric.
Speaker
-
Henri Orland
