Talks

Time series data contains useful information on the phase of a system. Here we propose the use of recurrent neural networks (LSTM) to learn and extract such information in order to classify phases and locate phase boundaries. We demonstrate this on a many-body localized model, and attempt to interpret the learned behavior by looking at individual LSTM cells. We also discuss the validity of the learned model and investigate its limits.

The causal structure of spacetimes containing fully evaporated black holes is considered from the perspective of Lorentzian geometry. The starting point is provided by theorems, due to Kodama, Geroch and Wald, that derive non-global hyperbolicity from a set of premises relating two partial Cauchy surfaces that are thought of as, respectively, lying before and after the evaporation. Here, we consider the Geroch-Wald theorem in the setting of a conformally embedded spacetime a la Penrose and show, under the assumption that complete null geodesics outside the black hole reach the boundary, that there exists visible incomplete null geodesics and that causal continuity fails.