Better Latent Spaces for Better Autoencoders

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.

Taming modeling uncertainties with Mass Unspecific Supervised Tagging

“…Generic taggers for multi-pronged jets can also be built by using representation learning, e.g. with an autoencoder [14][15][16][17][18]. Without the need of any signal assumption, but only using background (pseudo-)data, an unsupervised tagger can learn the background features in order to pinpoint outliers, i.e.…”

Challenges for Unsupervised Anomaly Detection in Particle Physics

“…[50]). Some methods that search for outliers rely on abstract representations to try to characterize the event space, such as the latent space of an autoencoder [20,48]. Others give the event space itself a geometric interpretation in terms of distances [55][56][57].…”

“…To do so, the autoencoder must establish a delicate balance in achieving a reconstruction fidelity which is accurate, but not too accurate. There are several cases where this is especially difficult, such as when the signal-to-background ratio S/B is small, when the dataset has certain topological properties [18], or when innate characteristics of the samples make the signal sample simpler than the background sample to reconstruct [46,48].…”

“…In order to understand what a VAE is learning, we study its latent space. In particular, we look at distance between events in VAE latent space (see [48,63] for other studies of VAE latent spaces in particle physics). Since we can think of the VAE anomaly score as a "distance" encoding how far any given event is from the background distribution, it is also natural to ask about the distances between individual events.…”

“…This estimation assumes Standard Normal priors for the likelihood of the latent data, as described in appendix A. There is a great deal of ongoing research into methods to improve the likelihood estimate by changing the latent space priors or improving the posterior approximations of the encoder[2,48,82,83].…”