Title: Optimal control in the space of probabilities: what do crowd dynamics and machine learning have to do with it?
Author: Maksim Staritsyn, SYSTEC, FEUP
DEEC I-105 November 16, 2022 – 4 pm
In this short talk, I will try to overview the line of research that we have focused on in the last several years. It deals with the optimal control theory for distributed systems in the space of probabilities. Such systems give an elegant mathematical representation of the dynamics of structurally homogeneous populations of interacting units (people, animals, mobile robots, elementary particles, or even neurons of artificial networks). To give you a basic intuition, I will start with toy models from crowd motion control and conclude with some recent (and even prospective) results related to few-shot learning of residual neural networks and quantum ensembles.