Autoencoders with latent spaces shaped like S^2, S^2×S^2, or RP^2, matched to the phase-space topology of the background, reduce spurious reconstruction errors and give a small but consistent anomaly-detection gain over flat latent spaces on simulated top-quark decays.
Better Latent Spaces for Better Autoencoders
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
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.
citation-role summary
background 1
citation-polarity summary
fields
hep-ph 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Enhancing anomaly detection with topology-aware autoencoders
Autoencoders with latent spaces shaped like S^2, S^2×S^2, or RP^2, matched to the phase-space topology of the background, reduce spurious reconstruction errors and give a small but consistent anomaly-detection gain over flat latent spaces on simulated top-quark decays.