Problem
Research questions and friction points this paper is trying to address.
simulation-based inference
bias
domain shift
signal fraction estimation
parameter inference
Many Wrongs Make a Right: Leveraging Biased Simulations Towards Unbiased Parameter Inference
🤖 AI Summary
This work addresses the bias in signal fraction estimation arising from modeling discrepancies between simulated and real data. To mitigate this issue, the authors propose a Template Adaptive Mixture Model that data-adaptively combines multiple biased simulations to dynamically correct for distributional shifts. The method innovatively integrates model selection, optimized feature representation, and statistical inference to achieve unbiased parameter estimation without requiring any true signal samples, while maintaining well-calibrated uncertainty quantification. Evaluations on both a Gaussian toy model and a semi-realistic di-Higgs measurement task demonstrate a significant reduction in estimation bias, confirming the approach’s effectiveness and robustness.
📝 Abstract
In particle physics, as in many areas of science, parameter inference relies on simulations to bridge the gap between theory and experiment. Recent developments in simulation-based inference have boosted the sensitivity of analyses; however, biases induced by simulation-data mismodeling can be difficult to control within standard inference pipelines. In this work, we propose a Template-Adapted Mixture Model to confront this problem in the context of signal fraction estimation: inferring the population proportion of signal in a mixed sample of signal and background, both of which follow arbitrarily complex distributions. We harness many biased simulations to perform data-driven estimates of each process distribution in the signal region, substantially reducing the bias on the signal fraction due to the domain shift between simulation and reality. We explore different methodological choices, including model selection, feature representation, and statistical method, and apply them to a Gaussian toy example and to a semi-realistic di-Higgs measurement. We find that the presented methods successfully leverage the biased simulations to provide estimates with well-calibrated uncertainties.
Innovation
Methods, ideas, or system contributions that make the work stand out.
simulation-based inference
domain shift
signal fraction estimation
mixture model
bias correction
🔎 Similar Papers
No similar papers found.
