Anomaly detection with convolutional Graph Neural Networks

Abstract

We devise an autoencoder based strategy to facilitate anomaly detection for boosted jets, employing Graph Neural Networks (GNNs) to do so. To overcome known limitations of GNN autoencoders, we design a symmetric decoder capable of simultaneously reconstructing edge features and node features. Focusing on latent space based discriminators, we find that such setups provide a promising avenue to isolate new physics and competing SM signatures from sensitivity-limiting QCD jet contributions. We demonstrate the flexibility and broad applicability of this approach using examples of W bosons, top quarks, and exotic hadronically-decaying exotic scalar bosons.

What's Anomalous in LHC Jets?

“…Based on these practical successes, ML-methods for anomaly detection at the LHC have generally received a lot of attention in the context of anomalous jets [10][11][12][13][14][15][16][17], anomalous events pointing to physics beyond the Standard Model [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35], or enhancing established search strategies [36][37][38][39][40][41][42]. They include a first ATLAS analysis [43], experimental validation of some of the methods [44,45], quantum machine learning [46], applications to heavy-ion collisions [47], the DarkMachines challenge [48], and the LHC Olympics 2020 community challenge [49,50].…”

Particle Graph Autoencoders and Differentiable, Learned Energy Mover's Distance

“…GNNs for anomaly detection in HEP have not yet been fully explored. However, recent work [6] develops an autoencoder-based strategy to facilitate anomaly detection for boosted jets, using a symmetric decoder that simultaneously reconstructs edge features and node features. Latent-space discriminators are used isolate W bosons, top quarks, and exotic hadronically-decaying exotic scalar bosons from QCD multijet background.…”

“…To address this, there has been a growing interest in employing unsupervised machine learning (ML) models that can search for new physics independent of underlying signature assumptions. For example, autoencoders, ML models that learn to map data down to a compressed encoding of its most salient features and then reverse such encodings back to their original form, have been employed for unsupervised anomaly detection [1][2][3][4][5][6]. Autoencoders learn to accurately reconstruct data similar to what is seen during its training; however, anomalous signals rare or absent in the training data may not be accurately reconstructed-a property that can be used to detect them.…”

Anomaly detection in high-energy physics using a quantum autoencoder

“…If the signal is kinematically sufficiently different from the background samples, the loss function or reconstruction error will be larger for signal than for background events. Such autoencoder can be augmented with convolutional neural networks [15,16], graph neural networks [17,18] or recurrent neural networks [19,20] on its outset, making it a very flexible anomaly detection method for a vast number of use-cases.…”

“…This better performance is particularly interesting as the CAE has O(1000) parameters compared to just O(10) for the QAE. The study indicates the possibility to study quantum latent spaces of high-energy collisions, in analogy to classical autoencoders [17,[44][45][46].…”