Regions of simulated glasses for binary classification about imminent rearrangement

Associated with "Information decomposition to identify relevant variation in complex systems with machine learning" (https://arxiv.org/abs/2307.04755)Kieran A. Murphy and Dani S. Bassett, 2023Derived from Richard et al. 2020 ("Predicting plasticity in disordered solids from structural indicators"), this dataset contains local neighborhoods in glasses prepared by either rapid or gradual quenching. Half of the neighborhoods are about to kick off a sudden rearrangement event, and the other half are sampled from elsewhere in the system (at the same instant in time) to create a balanced binary classification dataset. Further, the data has already been divided according to the train/val split used in the paper. Outside of the npz pickle (which contains particle positions, types, and the binary label for every even), the radial distribution functions g_AA(r) and g_AB(r) are included. These measure the density in radial bands around type A particles, averaged across the entire system, and were used to ground the information decomposition results in the paper. Code to process the glass data as we did in the paper can be found through our project page, https://distributed-information-bottleneck.github.io.