Equivariant Neural Networks

Last updated on Sep 10, 2023
Review

Developed a graph-based autoencoder (PGAE) for compression of and anomaly detection in Large Hadron Collider data. Wrote a review of deep learning models that are equivariant to physics-relevant group transformations for the UCSD group theory course. Led to developing a graph-based autoencoder equivariant to Lorentz group transformations as well (LGAE). LGAE outperformed CNN and GNN autoencoders for compression and anomaly detection tasks.

Publications

Zichun Hao, Raghav Kansal, Javier Duarte, Nadezda Chernyavskaya (2023). Lorentz group equivariant autoencoders. Eur. Phys. J. C.

PDF arXiv Cite Code Project Project DOI

Steven Tsan, Raghav Kansal, Anthony Aportela, Daniel Diaz, Javier Duarte, Sukanya Krishna, Farouk Mokhtar, Jean-Roch Vlimant, Maurizio Pierini (2021). Particle Graph Autoencoders and Differentiable, Learned Energy Mover's Distance. ML and the Physical Sciences Workshop @ NeurIPS 2021.

PDF arXiv Cite Project Project Poster

Talks and Posters

UCI Particle/Astro ML Seminar, Generative transformers and how to evaluate them (+ Lorentz-equivariant networks)
Jun 27, 2023 UC Irvine Invited Talk
Project Project Project Slides