SciPost logo

Better latent spaces for better autoencoders

Barry M. Dillon, Tilman Plehn, Christof Sauer, Peter Sorrenson

SciPost Phys. 11, 061 (2021) · published 17 September 2021

Abstract

Autoencoders as tools behind anomaly searches at the LHC have the structural problem that they only work in one direction, extracting jets with higher complexity but not the other way around. To address this, we derive classifiers from the latent space of (variational) autoencoders, specifically in Gaussian mixture and Dirichlet latent spaces. In particular, the Dirichlet setup solves the problem and improves both the performance and the interpretability of the networks.


Authors / Affiliation: mappings to Contributors and Organizations

See all Organizations.
Funders for the research work leading to this publication