Hybrid Neural Simulation-Based Inference for Robust Applications and Limited-Budget Scenarios
Authors: Sean Benevedes, Mani Dehghan, Aishik Ghosh, Tae Hyoun Park
Organizations: School of Physics, Georgia Institute of Technology, Atlanta, GA 30332, USA · School of Physics, Georgia Institute of Technology, Atlanta, GA 30332, USA and Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA · Max Planck Institute for Physics, Boltzmannstr. 8, 85748 Garching, Germany
We develop two hybrid techniques that approach the performance of neural simulation-based inference (NSBI) analyses while substantially reducing the computational cost of inference and preserving some or all of the reliability guarantees of parametric methods. The first approach is broadly applicable, while the second is tailored to a class of particle physics analyses that admit a semi-parametric NSBI formulation. With only a modest compromise in raw sensitivity, these methods represent an important step toward computationally efficient NSBI in offline analyses and also open the door to the exploration of trigger-level applications in the future. Based on our comparison studies, we recommend the use of our first approach, Latent Categories, for robust and efficient inference.
Figures & tables
Figure 1: Schematic diagram of the training and inference pipelines for the Latent Categories approach. Black indicates components of the architecture present at both training and inference time. Orange indicates training-specific elements, showing how the encoder is trained as a multiclassifier, and green shows the inference pipeline.
Figure 2: Schematic diagram of the training and inference pipelines for the Mixture of Summary Statistics approach. As with Fig. 1 , orange shows how the model is trained and green shows how it is used for inference.
Figure 3: Sensitivity of LC, NSBI, and histogram analyses on the Gaussian example dataset. Narrower curves indicate better sensitivity. The true value of the parameter is shown with a vertical dotted line, and the 1σ and 2σ thresholds are shown with the horizontal dashed and dotted lines.
Figure 4: Sensitivity of LC, MSS, NSBI, and histogram analyses on the off-shell Higgs dataset. Narrower curves indicate better sensitivity. The true value of the parameter is shown with a vertical dotted line, and the 1σ and 2σ thresholds are shown with the horizontal dashed and dotted lines.
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 5: Expected occupancy of each Latent Categories bin for the nominal parameter value θ=0 and the boundary values θ=±1 in the Gaussian case study. The categories, which are intrinsically unordered, are sorted here in order of descending occupancy in the θ=0 case.
Figure 6: Expected occupancy of each Latent Categories bin for the nominal parameter value μ=1 and the boundary values μ=0,4 in the physics case study. The categories, which are intrinsically unordered, are sorted here in order of descending occupancy in the μ=1 case.
Figure 7: A plot of observed versus expected coverage for the Gaussian case study at θ=0 . The dotted black line shows nominal coverage, the dashed gray line shows the coverage attained using the exact likelihood, and each of the solid lines corresponds to coverage of one of the procedures considered in the main text. The shading corresponds to the 1σ uncertainty in the observed coverage corresponding to the finite number of pseudo-experiments.
Figure 8: A plot of observed versus expected coverage for the physics case study at μ=10 . The dotted black line shows nominal coverage, the dashed gray line shows the coverage attained using the exact likelihood, and each of the solid lines corresponds to coverage of one of the procedures considered in the main text (except for the “MSS 3 bins” procedure, which is only considered in this appendix). The shading corresponds to the 1σ uncertainty in the observed coverage corresponding to the finite number of pseudo-experiments.
Neural simulation-based inference (SBI) has been widely successful in inferring a relatively small number of interpretable parameters from potentially high-dimensional observations, such as images or time series. Accordingly, representation learning in SBI has focused almost exclusively on compressing the observations used to condition the posterior. More recently, however, SBI has begun to target increasingly high-dimensional parameter spaces, raising the complementary question of whether the inference target itself should be compressed. Our answer is a practical merger of SBI and latent generative modeling, which learns a low-dimensional representation of the simulator parameters, performs posterior inference directly in this latent space, and maps posterior samples back to the original parameter space. We characterize the conditions under which latent-space inference recovers the desired target posterior and systematically study its empirical trade-offs. Across four case studies and three generative families, we compare latent and standard estimators while controlling for network capacity, regularization, optimization, and training compute. At matched training compute, latent-space inference achieves accuracy and marginal calibration comparable to direct target-space inference while sampling up to more than an order of magnitude faster.
Lars Kühmichel, Stefan T. Radev, Bhanu Prasanna Koppolu +3
Department of Statistics, TU Dortmund University, Germany · Department of Cognitive Science, Rensselaer Polytechnic Institute
Neural simulation-based inference enables parameter estimation for complex models, but typically requires the user to specify a simulator encoding a fixed model structure. We present a framework for joint model selection and parameter estimation that combines large language models for program synthesis with neural simulation-based inference. Given a natural language description of the system and data under investigation, an LLM proposes candidate simulator programs which are iteratively refined via feedback-driven mutation and evaluated using neural density estimation. The approach enables simulation-based inference over a pool of models, not just parameters within a fixed model. On benchmarks spanning deterministic dynamics, stochastic epidemic models, and dark matter substructure inference from gravitational-lensing images, the method identifies plausible model families from open-ended prompts, with accuracy that reflects the information content of the data and identifiability of candidate models.
Siddharth Mishra-Sharma
Faculty of Computing & Data Sciences Boston University · Boston University
Simulation-based inference (SBI) with machine learning is an increasingly important tool for solving inverse problems in science and engineering, including parameter inference and the inversion of detector effects. We provide an overview of the Bayesian and frequentist statistical frameworks, describe how machine-learning-based SBI methods, such as neural posterior estimation and neural likelihood estimation, can be used for parameter estimation within these frameworks, and show that the same methods can also be applied to Empirical Bayes or unfolding tasks. We also discuss how to validate inference results and the limitations of SBI with machine learning.
Maximilian Dax, Theo Heimel, Gilles Louppe
ELLIS Institute Tübingen, Max Planck Institute for Intelligent Systems, Tübingen AI Center, Tübingen, Germany · CP3, Université catholique de Louvain, Louvain-la-Neuve, Belgium · Montefiore Institute, University of Liège, Belgium