Nonlocal Mean Field Schr\"{o}dinger Bridge with Learned Interactions
Researchers have developed a new method for the Mean-Field Schrödinger Bridge problem, which models interacting particle systems. Their approach uses neural networks to approximate nonlocal interactions, reducing computational costs from quadratic to linear with population size. This technique was demonstrated to be effective in numerical experiments for navigation and opinion-dynamics tasks, significantly cutting down training time while maintaining trajectory accuracy. AI