JUNIPR: a Framework for Unsupervised Machine Learning in Particle Physics

Abstract

In applications of machine learning to particle physics, a persistent challenge is how to go beyond discrimination to learn about the underlying physics. To this end, a powerful tool would be a framework for unsupervised learning, where the machine learns the intricate high-dimensional contours of the data upon which it is trained, without reference to pre-established labels. In order to approach such a complex task, an unsupervised network must be structured intelligently, based on a qualitative understanding of the data. In this paper, we scaffold the neural network's architecture around a leading-order model of the physics underlying the data. In addition to making unsupervised learning tractable, this design actually alleviates existing tensions between performance and interpretability. We call the framework JUNIPR: "Jets from UNsupervised Interpretable PRobabilistic models". In this approach, the set of particle momenta composing a jet are clustered into a binary tree that the neural network examines sequentially. Training is unsupervised and unrestricted: the network could decide that the data bears little correspondence to the chosen tree structure. However, when there is a correspondence, the network's output along the tree has a direct physical interpretation. JUNIPR models can perform discrimination tasks, through the statistically optimal likelihood-ratio test, and they permit visualizations of discrimination power at each branching in a jet's tree. Additionally, JUNIPR models provide a probability distribution from which events can be drawn, providing a data-driven Monte Carlo generator. As a third application, JUNIPR models can reweight events from one (e.g. simulated) data set to agree with distributions from another (e.g. experimental) data set.

arXiv comment37 pages, 24 figures

References in corpus (28)

• PYTHIA 6.4 Physics and Manual
• PYTHIA 6.2 Physics and Manual
• The anti-k_t jet clustering algorithm
• FastJet user manual
• Successive Combination Jet Algorithm For Hadron Collisions
• Better Jet Clustering Algorithms
• Identifying Boosted Objects with N-subjettiness
• CaloGAN: Simulating 3D High Energy Particle Showers in Multi-Layer Electromagnetic Calorimeters with Generative Adversarial Networks
• Classification without labels: Learning from mixed samples in high energy physics
• Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics Synthesis
• Deep learning in color: towards automated quark/gluon jet discrimination
• Accelerating Science with Generative Adversarial Networks: An Application to 3D Particle Showers in Multi-Layer Calorimeters
• Deep-learning Top Taggers or The End of QCD?
• Jet Flavor Classification in High-Energy Physics with Deep Neural Networks
• Seeing in Color: Jet Superstructure
• Jet-Images: Computer Vision Inspired Techniques for Jet Tagging
• Finding physics signals with shower deconstruction
• Hard-Soft-Collinear Factorization to All Orders
• Pileup Mitigation with Machine Learning (PUMML)
• Multivariate discrimination and the Higgs + W/Z search
• How Much Information is in a Jet?
• Novel Jet Observables from Machine Learning
• Finding physics signals with event deconstruction
• Qjets: A Non-Deterministic Approach to Tree-Based Jet Substructure
• (Machine) Learning to Do More with Less
• An on-shell approach to factorization
• Jet Sampling: Improving Event Reconstruction through Multiple Interpretations
• Fuzzy Jets

Cited by in corpus (83)

• The frontier of simulation-based inference
• ParticleNet: Jet Tagging via Particle Clouds
• Jet Substructure at the Large Hadron Collider: A Review of Recent Advances in Theory and Machine Learning
• Anomaly Detection with Density Estimation
• Searching for New Physics with Deep Autoencoders
• The Machine Learning Landscape of Top Taggers
• How to GAN LHC Events
• OmniFold: A Method to Simultaneously Unfold All Observables
• QCD or What?
• Getting High: High Fidelity Simulation of High Granularity Calorimeters with High Speed
• Novelty Detection Meets Collider Physics
• Adversarially-trained autoencoders for robust unsupervised new physics searches
• The Metric Space of Collider Events
• Anomaly detection in high-energy physics using a quantum autoencoder
• Machine Learning and LHC Event Generation
• Uncovering latent jet substructure
• Unsupervised Manifold Clustering of Topological Phononics
• Invertible Networks or Partons to Detector and Back Again
• Neural Networks for Full Phase-space Reweighting and Parameter Tuning
• A comprehensive review of Quantum Machine Learning: from NISQ to Fault Tolerance
• How to GAN away Detector Effects
• Understanding Event-Generation Networks via Uncertainties
• Event Generators for High-Energy Physics Experiments
• GANplifying Event Samples
• Generative Networks for Precision Enthusiasts
• A Neural Resampler for Monte Carlo Reweighting with Preserved Uncertainties
• Fast Point Cloud Generation with Diffusion Models in High Energy Physics
• Targeting Multi-Loop Integrals with Neural Networks
• A guide for deploying Deep Learning in LHC searches: How to achieve optimality and account for uncertainty
• Deep-Learning Jets with Uncertainties and More
• Mapping Machine-Learned Physics into a Human-Readable Space
• Interpretable Deep Learning for Two-Prong Jet Classification with Jet Spectra
• Exploring SMEFT in VH with Machine Learning
• Symmetries, Safety, and Self-Supervision
• Per-Object Systematics using Deep-Learned Calibration
• Learning to Simulate High Energy Particle Collisions from Unlabeled Data
• E Pluribus Unum Ex Machina: Learning from Many Collider Events at Once
• How to Understand Limitations of Generative Networks
• Unsupervised Hadronic SUEP at the LHC
• Two Invertible Networks for the Matrix Element Method
• Automating the Construction of Jet Observables with Machine Learning
• Exploring anomalous couplings in Higgs boson pair production through shape analysis
• Binary JUNIPR: an interpretable probabilistic model for discrimination
• Neural Network-based Top Tagger with Two-Point Energy Correlations and Geometry of Soft Emissions
• Calculating the primary Lund Jet Plane density
• Learning the language of QCD jets with transformers
• The MadNIS Reloaded
• Machine Learning in High Energy Physics: A review of heavy-flavor jet tagging at the LHC
• Machine learning-based jet and event classification at the Electron-Ion Collider with applications to hadron structure and spin physics
• Explainable machine learning of the underlying physics of high-energy particle collisions
• High-dimensional and Permutation Invariant Anomaly Detection
• How to GAN Event Subtraction
• CaloDREAM -- Detector Response Emulation via Attentive flow Matching
• Learning Likelihood Ratios with Neural Network Classifiers
• CapsNets Continuing the Convolutional Quest
• Jet grooming through reinforcement learning
• Precision-Machine Learning for the Matrix Element Method
• ELSA -- Enhanced latent spaces for improved collider simulations
• Reconstructing boosted Higgs jets from event image segmentation
• Quantum Generative Adversarial Networks For Anomaly Detection In High Energy Physics
• Normalizing Flows for High-Dimensional Detector Simulations
• Matrix Element Regression with Deep Neural Networks -- breaking the CPU barrier
• Convolutional L2LFlows: Generating Accurate Showers in Highly Granular Calorimeters Using Convolutional Normalizing Flows
• Differentiable MadNIS-Lite
• Learning to Inflate
• Parameter Inference from Event Ensembles and the Top-Quark Mass
• Learning Physics at Future $e^{-}{e}^{+}$ Colliders with Machine
• A Comparison Between Invariant and Equivariant Classical and Quantum Graph Neural Networks
• Hypergraphs in LHC Phenomenology -- The Next Frontier of IRC-Safe Feature Extraction
• Deep Learning Jet Substructure from Two-Particle Correlation
• Automatic detection of boosted Higgs boson and top quark jets in an event image
• Unsupervised and lightly supervised learning in particle physics
• How to GAN Event Unweighting
• Identifying Heavy-Flavor Jets Using Vectors of Locally Aggregated Descriptors
• Reconstructing axion-like particles from beam dumps with simulation-based inference
• Classify the Higgs decays with the PFN and ParticleNet at electron-positron colliders
• How much joint resummation do we need?
• Extrapolating Jet Radiation with Autoregressive Transformers
• BitHEP -- The Limits of Low-Precision ML in HEP
• Introduction and analysis of a method for the investigation of QCD-like tree data
• Jet Reconstruction with Mamba Networks in Collider Events
• Observable Optimization for Precision Theory: Machine Learning Energy Correlators
• Power Counting Energy Flow Polynomials