DisCo Fever: Robust Networks Through Distance Correlation

Abstract

While deep learning has proven to be extremely successful at supervised classification tasks at the LHC and beyond, for practical applications, raw classification accuracy is often not the only consideration. One crucial issue is the stability of network predictions, either versus changes of individual features of the input data, or against systematic perturbations. We present a new method based on a novel application of "distance correlation" (DisCo), a measure quantifying non-linear correlations, that achieves equal performance to state-of-the-art adversarial decorrelation networks but is much simpler and more stable to train. To demonstrate the effectiveness of our method, we carefully recast a recent ATLAS study of decorrelation methods as applied to boosted, hadronic W-tagging. We also show the feasibility of DisCo regularization for more powerful convolutional neural networks, as well as for the problem of hadronic top tagging.

9 pages, v2: essentially the journal version (refs added, typos fixed, minor improvements)

References in corpus (9)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Measuring and testing dependence by correlation of distances
• Brownian distance covariance
• Power Counting to Better Jet Observables
• Learning in Implicit Generative Models
• Search for direct top squark pair production in events with one lepton, jets, and missing transverse momentum at 13 TeV with the CMS experiment
• Lorentz Boost Networks: Autonomous Physics-Inspired Feature Engineering
• Long Short-Term Memory (LSTM) networks with jet constituents for boosted top tagging at the LHC
• Jet-Parton Assignment in ttH Events using Deep Learning

Cited by in corpus (42)

• Anomaly Detection with Density Estimation
• Dijet resonance search with weak supervision using $\sqrt{s}=13$ TeV $pp$ collisions in the ATLAS detector
• Tag N' Train: A Technique to Train Improved Classifiers on Unlabeled Data
• Variational Autoencoders for Anomalous Jet Tagging
• Simulation-Assisted Decorrelation for Resonant Anomaly Detection
• Bump Hunting in Latent Space
• Machine Learning for Anomaly Detection in Particle Physics
• ABCDisCo: Automating the ABCD Method with Machine Learning
• Uncertainty Aware Learning for High Energy Physics
• Quasi Anomalous Knowledge: Searching for new physics with embedded knowledge
• Online-compatible Unsupervised Non-resonant Anomaly Detection
• Equivariant Energy Flow Networks for Jet Tagging
• A neural simulation-based inference approach for characterizing the Galactic Center $\gamma$-ray excess
• The LHC Olympics 2020: A Community Challenge for Anomaly Detection in High Energy Physics
• Dark sector searches with the CMS experiment
• A Cautionary Tale of Decorrelating Theory Uncertainties
• Search for R-parity violating supersymmetry in a final state containing leptons and many jets with the ATLAS experiment using $\sqrt{s}=13$ TeV proton-proton collision data
• A Holistic Approach to Predicting Top Quark Kinematic Properties with the Covariant Particle Transformer
• Optimal statistical inference in the presence of systematic uncertainties using neural network optimization based on binned Poisson likelihoods with nuisance parameters
• Search for heavy Higgs bosons with flavour-violating couplings in multi-lepton plus $b$-jets final states in $pp$ collisions at 13 TeV with the ATLAS detector
• Enhancing searches for resonances with machine learning and moment decomposition
• GAN-AE : An anomaly detection algorithm for New Physics search in LHC data
• Search for long-lived particles using displaced vertices and missing transverse momentum in proton-proton collisions at $\sqrt{s}$ = 13 TeV
• Bias and Priors in Machine Learning Calibrations for High Energy Physics
• Interpretable Uncertainty Quantification in AI for HEP
• Meta-learning and data augmentation for mass-generalised jet taggers
• Dealing with Nuisance Parameters using Machine Learning in High Energy Physics: a Review
• Invariant Representation Driven Neural Classifier for Anti-QCD Jet Tagging
• Deeply Learned Preselection of Higgs Dijet Decays at Future Lepton Colliders
• Robust Distance Covariance
• Decorrelation using Optimal Transport
• PAIReD jet: A multi-pronged resonance tagging strategy across all Lorentz boosts
• Riemannian Data preprocessing in Machine Learning to focus on QCD color structure
• Observation of the Singly Cabibbo-Suppressed Decay ${\Lambda }_c^{+}\to p{\pi }^0$
• Evidence for the dimuon decay of the Higgs boson in pp collisions with the ATLAS detector
• Designing Observables for Measurements with Deep Learning
• A search for dark matter produced in association with a dark Higgs boson decaying into a Higgs boson pair in 3$b$ or 4$b$ final states using $pp$ collisions at $\sqrt{s}=13$ TeV with the ATLAS detector
• Image-Based Jet Analysis
• Robust Anomaly Detection for Particle Physics Using Multi-Background Representation Learning
• Enhancing generalization in high energy physics using white-box adversarial attacks
• Reweighting and Analysing Event Generator Systematics by Neural Networks on High-Level Features
• Search for the radiative leptonic decay $D^{+}\to \gamma e^{+}{\nu }_e$ using Deep Learning