Discriminative versus Generative Approaches to Simulation-based Inference

2025-03-11

AI Summary

This study systematically compares neural likelihood estimation (NLE) and neural likelihood ratio estimation (NRE)—two simulation-based inference (SBI) methods—for high-dimensional, unbinned parameter estimation in particle and nuclear physics. Under a unified experimental setup, we conduct the first fair benchmark of both methods on a Gaussian benchmark model and the realistic FAIR Universe Higgs simulation dataset. Results demonstrate that NRE achieves significantly higher estimation accuracy and precision than NLE, particularly under low signal-to-noise conditions and in high-dimensional observation spaces; in contrast, NLE exhibits substantial training variance, necessitating ensemble strategies for robustness. The findings highlight the intrinsic advantages of discriminative likelihood-ratio modeling for physics inference, revealing its superior statistical efficiency and stability. This work provides empirical evidence and practical guidance for selecting appropriate SBI methods in high-energy physics applications.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Probabilistic Inference
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Most of the fundamental, emergent, and phenomenological parameters of particle and nuclear physics are determined through parametric template fits. Simulations are used to populate histograms which are then matched to data. This approach is inherently lossy, since histograms are binned and low-dimensional. Deep learning has enabled unbinned and high-dimensional parameter estimation through neural likelihiood(-ratio) estimation. We compare two approaches for neural simulation-based inference (NSBI): one based on discriminative learning (classification) and one based on generative modeling. These two approaches are directly evaluated on the same datasets, with a similar level of hyperparameter optimization in both cases. In addition to a Gaussian dataset, we study NSBI using a Higgs boson dataset from the FAIR Universe Challenge. We find that both the direct likelihood and likelihood ratio estimation are able to effectively extract parameters with reasonable uncertainties. For the numerical examples and within the set of hyperparameters studied, we found that the likelihood ratio method is more accurate and/or precise. Both methods have a significant spread from the network training and would require ensembling or other mitigation strategies in practice.

Problem

• Comparing discriminative and generative approaches for simulation-based inference.
• Evaluating neural methods for high-dimensional parameter estimation in physics.
• Assessing accuracy and precision of likelihood ratio estimation in NSBI.

Innovation

• Neural likelihood estimation for unbinned analysis
• Comparison of discriminative and generative NSBI methods
• Likelihood ratio method enhances accuracy and precision