ML for Fast Simulations

Developed a graph-based generative adversarial network, MPGAN, which proved to be state-of-the-art at simulating particle collisions. Developed as well the faster attention-based generative adversarial particle transformer (GAPT), using set transformers, and efficient and sensitive two-sample goodness-of-fit tests for validating fast simulations. Working now on extending to conditional generation and application to detector data.