Improving Variational Autoencoders for New Physics Detection at the LHC With Normalizing Flows
Pratik Jawahar, T. K. Aarrestad, N. Chernyavskaya, M. Pierini, Kinga A. Wozniak, J. Ngadiuba, J. Duarte, Steven Tsan
Abstract
We investigate how to improve new physics detection strategies exploiting variational autoencoders and normalizing flows for anomaly detection at the Large Hadron Collider. As a working example, we consider the DarkMachines challenge dataset. We show how different design choices (e.g., event representations, anomaly score definitions, network architectures) affect the result on specific benchmark new physics models. Once a baseline is established, we discuss how to improve the anomaly detection accuracy by exploiting normalizing flow layers in the latent space of the variational autoencoder.
Topics & Concepts
AutoencoderAnomaly detectionLarge Hadron ColliderAnomaly (physics)Benchmark (surveying)Event (particle physics)Particle physicsFlow (mathematics)Physics beyond the Standard ModelComputer sciencePhysicsArtificial intelligenceMachine learningAlgorithmArtificial neural networkQuantum mechanicsMechanicsGeodesyGeographyParticle physics theoretical and experimental studiesAnomaly Detection Techniques and ApplicationsComputational Physics and Python Applications