Prof. Gabriel Peyré, a CNRS research director and professor at Ecole Normale Supérieure (France) gave an iTHEMS-AIP joint colloquium entitled "Scaling Optimal Transport for High dimensional Learning" on January 24th. He started his talk with the old Monge's problem and moved to modern concept of the optimal transport by Kantorovich with historical stories behind it. He showed us that there have been many efforts of various mathematicians, economists, and physicists behind the developments of the optimal transport theory. In the later part of his talk, Prof. Gabriel Peyré focused on optimal transport problems in extremely high dimensions looking ahead various potential scientific applications like imaging, natural language processing, and biology. He introduced the key concept, entoropic term, with the historical quotation from Schrödinger, and showed us how it enables approximately solving optimal transport problems in high dimensions. He made his talk with a lot of beautiful figures and detailed explanations. We have many discussions during/after the talk, and enjoyed the colloquium.

Reported by Akinori Tanaka

関連イベント