Equivariant and Physics-Informed ML

Developed graph-based and Lorentz-equivariant models to better suit our high energy physics data. Our Lorentz-group autoencoder (LGAE) outperforms graph and convolutional networks on jet compression and anomaly detection tasks. Latest work published at EPJC.

Developed several approaches for finding rare particle collisions, including GNNs, Lorentz-equivariant networks, and multi-variate goodness-of-fit tests.

A novel physics-informed self-supervised pretraining strategy for foundation models in physics. SOTA results on transfer learning between simulations and real data.

Fast simulations which can accurately model jet substructure are will be of utmost importance for boosted jet analyses at the HL-LHC. …

Accurate detector simulations are key components of any measurement or search for new physics. Due to their stochastic nature, ML-based …

With the increase in luminosity and detector granularity, simulation will be a significant computational challenge in the HL-LHC. To …

Developed an auto-encoder model equivariant to Lorentz transformations of the input. We find it outperforms graph and convolutional neural networks on jet reconstruction and anomaly detection tasks.

With the increase in luminosity and detector granularity, simulation will be a significant computational challenge in the HL-LHC. To …