Dear or alive: Distinguishing active from passive particles using supervised learning

A longstanding open question in the field of dense disordered matter is how precisely the structure and the dynamics are related to each other. With the advent of machine learning, it is now becoming possible to agnostically identify structure-dynamics correlations that would be virtually impossible to see with the naked eye. In this work we employ a supervised learning approach to study glassy mixtures that are composed of active and passive Brownian particles. Based on local structural order parameters obtained from a single snapshot, our neural network is able to predict with almost 100% accuracy which particles are active and which ones are not. Hence, this machine learning model can identify distinct dynamical single-particle properties on purely structural grounds. Ultimately, these efforts might also find relevance in the context of biological active glasses such as confluent cell layers, where subtle changes in the microstructure can hint at pathological changes in the cell dynamics.