Mapping Machine-Learned Physics into a Human-Readable Space

Abstract

We present a technique for translating a black-box machine-learned classifier operating on a high-dimensional input space into a small set of human-interpretable observables that can be combined to make the same classification decisions. We iteratively select these observables from a large space of high-level discriminants by finding those with the highest decision similarity relative to the black box, quantified via a metric we introduce that evaluates the relative ordering of pairs of inputs. Successive iterations focus only on the subset of input pairs that are misordered by the current set of observables. This method enables simplification of the machine-learning strategy, interpretation of the results in terms of well-understood physical concepts, validation of the physical model, and the potential for new insights into the nature of the problem itself. As a demonstration, we apply our approach to the benchmark task of jet classification in collider physics, where a convolutional neural network acting on calorimeter jet images outperforms a set of six well-known jet substructure observables. Our method maps the convolutional neural network into a set of observables called energy flow polynomials, and it closes the performance gap by identifying a class of observables with an interesting physical interpretation that has been previously overlooked in the jet substructure literature.

20 pages, 7 figures, 3 tables

References in corpus (25)

• PYTHIA 6.4 Physics and Manual
• Jet substructure as a new Higgs search channel at the LHC
• Observation of the diphoton decay of the Higgs boson and measurement of its properties
• Top-tagging: A Method for Identifying Boosted Hadronic Tops
• Boosted objects: a probe of beyond the Standard Model physics
• Classification without labels: Learning from mixed samples in high energy physics
• Power Counting to Better Jet Observables
• Gaining (Mutual) Information about Quark/Gluon Discrimination
• Jet Flavor Classification in High-Energy Physics with Deep Neural Networks
• New Angles on Energy Correlation Functions
• Parton Shower Uncertainties in Jet Substructure Analyses with Deep Neural Networks
• Identification techniques for highly boosted W bosons that decay into hadrons
• Enhanced Higgs to ${\tau }^{+}{\tau }^{-}$ Searches with Deep Learning
• Jet Substructure Without Trees
• Reconstructing Sparticle Mass Spectra using Hadronic Decays
• A neural network clustering algorithm for the ATLAS silicon pixel detector
• W-jet Tagging: Optimizing the Identification of Boosted Hadronically-Decaying W Bosons
• Search for a Higgs boson in the decay channel H to ZZ(*) to q qbar l-l+ in pp collisions at sqrt(s) = 7 TeV
• Lorentz Group Equivariant Neural Network for Particle Physics
• Quark-Gluon tagging with Shower Deconstruction: Unearthing dark matter and Higgs couplings
• Towards Machine Learning Analytics for Jet Substructure
• Measurement of the cross-section of high transverse momentum vector bosons reconstructed as single jets and studies of jet substructure in $pp$ collisions at $\sqrt{s}$ = 7 TeV with the ATLAS detector
• An Optimal Observable for Color Singlet Identification
• Cutting Multiparticle Correlators Down to Size
• Classification of Energy Flow Observables in Narrow Jets

Cited by in corpus (28)

• Equivariant Energy Flow Networks for Jet Tagging
• How to Understand Limitations of Generative Networks
• Energy-weighted Message Passing: an infra-red and collinear safe graph neural network algorithm
• Deep learning jet modifications in heavy-ion collisions
• Explainable machine learning of the underlying physics of high-energy particle collisions
• Triggering on Emerging Jets
• Combine and Conquer: Event Reconstruction with Bayesian Ensemble Neural Networks
• Learning to Identify Electrons
• Resolving Extreme Jet Substructure
• Learning to Identify Semi-Visible Jets
• Quantum-inspired event reconstruction with Tensor Networks: Matrix Product States
• Is infrared-collinear safe information all you need for jet classification?
• Learning to Isolate Muons
• Hypergraphs in LHC Phenomenology -- The Next Frontier of IRC-Safe Feature Extraction
• Interpretable deep learning models for the inference and classification of LHC data
• TopicFlow: Disentangling quark and gluon jets with normalizing flows
• Explainable AI for ML jet taggers using expert variables and layerwise relevance propagation
• Riemannian Data preprocessing in Machine Learning to focus on QCD color structure
• Top-philic Machine Learning
• Foundations of automatic feature extraction at LHC--point clouds and graphs
• NCoder -- A Quantum Field Theory approach to encoding data
• Deep-Learned Event Variables for Collider Phenomenology
• Designing Observables for Quantum Interference in Jets at Subleading Color
• Applying Machine Learning Techniques To Intermediate-Length Cascade Decays
• Creating Simple, Interpretable Anomaly Detectors for New Physics in Jet Substructure
• Towards an Interpretable Data-driven Trigger System for High-throughput Physics Facilities
• Learnable cut flow for high energy physics
• Power Counting Energy Flow Polynomials