Approximating Likelihood Ratios with Calibrated Discriminative Classifiers

2015/06/06 by K. Cranmer, Cranmer, Kyle, Juan Pavez +3 · 36 citations
Computer Science · #Gaussian Processes and Bayesian Inference #Algorithms and Data Compression #Bayesian Methods and Mixture Models

paper · pdf · doi:10.48550/arxiv.1506.02169

Abstract

In many fields of science, generalized likelihood ratio tests are established tools for statistical inference. At the same time, it has become increasingly common that a simulator (or generative model) is used to describe complex processes that tie parameters θ of an underlying theory and measurement apparatus to high-dimensional observations x∈ ℝ^p. However, simulator often do not provide a way to evaluate the likelihood function for a given observation x, which motivates a new class of likelihood-free inference algorithms. In this paper, we show that likelihood ratios are invariant under a specific class of dimensionality reduction maps ℝ^p ↦ ℝ. As a direct consequence, we show that discriminative classifiers can be used to approximate the generalized likelihood ratio statistic when only a generative model for the data is available. This leads to a new machine learning-based approach to likelihood-free inference that is complementary to Approximate Bayesian Computation, and which does not require a prior on the model parameters. Experimental results on artificial problems with known exact likelihoods illustrate the potential of the proposed method.

Citations

• Observation of a new particle in the search for the Standard Model Higgs boson with the ATLAS detector at the LHC

Cited by

• Minimum Distance Summaries for Robust Neural Posterior Estimation
• Quasiprobabilistic Density Ratio Estimation with a Reverse Engineered Classification Loss Function
• ColliderML: The First Release of an OpenDataDetector High-Luminosity Physics Benchmark Dataset
• Global fits and the search for new physics: past, present and future
• PolySwyft: sequential simulation-based nested sampling
• Machine Learning-based Unfolding for Cross Section Measurements in the Presence of Nuisance Parameters
• Electoral David vs Goliath: How does the Spatial Concentration of Electors affect District-based Elections?
• Machine learning and the physical sciences
• Neural Fake Factor Estimation Using Data-Based Inference
• Likelihood-free MCMC with Amortized Approximate Ratio Estimators
• Improvements to Inference Compilation for Probabilistic Programming in Large-Scale Scientific Simulators
• Benchmarking Simulation-Based Inference
• Fair Universe Higgs Uncertainty Challenge
• Simulation-efficient marginal posterior estimation with swyft: stop wasting your precious time
• Learning in Implicit Generative Models
• CP4SBI: Local Conformal Calibration of Credible Sets in Simulation-Based Inference
• Misspecification-robust amortised simulation-based inference using variational methods
• Unbinning global LHC analyses
• Searching for HWW Anomalous Couplings with Simulation-Based Inference
• Higgs Signal Strength Estimation with Machine Learning under Systematic Uncertainties
• Hierarchical Implicit Models and Likelihood-Free Variational Inference
• Mining gold from implicit models to improve likelihood-free inference
• Observable Optimization for Precision Theory: Machine Learning Energy Correlators
• Arbitrary Marginal Neural Ratio Estimation for Simulation-based\n Inference
• Simulation-based inference with deep ensembles: Evaluating calibration uncertainty and detecting model misspecification
• Trustworthy scientific inference with generative models
• Robust Simulation Based Inference
• How to embed any likelihood into SBI: Application to Planck + Stage IV galaxy surveys and Dynamical Dark Energy
• Analysis note: measurement of thrust in e⁺e^- collisions at √(s) = 91 GeV with archived ALEPH data
• On Focusing Statistical Power for Searches and Measurements in Particle Physics
• Flows for simultaneous manifold learning and density estimation
• Recent Advances in Simulation-based Inference for Gravitational Wave Data Analysis
• Diagnosing model misspecification and performing generalized Bayes'\n updates via probabilistic classifiers
• Neural simulation-based inference of the Higgs trilinear self-coupling via off-shell Higgs production
• Dealing with Nuisance Parameters using Machine Learning in High Energy Physics: a Review
• Generalized Energy Based Models

Related

• Likelihood-free inference via classification
• Classification and Bayesian Optimization for Likelihood-Free Inference
• Bayesian Optimization for Likelihood-Free Inference of Simulator-Based Statistical Models
• Likelihood-free inference by ratio estimation
• Likelihood-Free Inference with Generative Neural Networks via Scoring Rule Minimization