Abstract

Machine learning (ML) techniques are changing Science and Engineering by offering ways to reconsider complex problems once thought intractable because of the curse of dimensionality. In this talk, I will discuss the impact of ML on applications from active matter. Specifically I will show how physics-informed neural networks (PINNs) can be used to analyze first-order phase transitions in non-equilibrium systems, and how advances in generative modeling can be leveraged to characterize the breakup of time-reversal symmetry (TRS) and calculate entropy production rates (EPRs) in active systems. I will also discuss how the standard modus operandi of ML must be adapted in the context of such applications when they come with models and no data (as opposed to data and no models), and thereby require to use active learning strategies for data acquisition.