Solving Simulation Systematics in and with AI/ML

TL;DR

This paper addresses the challenge of systematic errors in AI/ML systems trained on simulated data for real detector analysis, proposing methods to quantify, minimize, and incorporate simulation uncertainties using AI/ML techniques.

Contribution

It introduces novel approaches to estimate and reduce simulation-related systematic errors in AI/ML models, including integrating simulation into training for meaningful parameter inference.

Findings

Methods to quantify simulation systematic uncertainties

Techniques to minimize simulation-related errors

Strategies to incorporate simulation into AI/ML training

Abstract

Training an AI/ML system on simulated data while using that system to infer on data from real detectors introduces a systematic error which is difficult to estimate and in many analyses is simply not confronted. It is crucial to minimize and to quantitatively estimate the uncertainties in such analysis and do so with a precision and accuracy that matches those that AI/ML techniques bring. Here we highlight the need to confront this class of systematic error, discuss conventional ways to estimate it and describe ways to quantify and to minimize the uncertainty using methods which are themselves based on the power of AI/ML. We also describe methods to introduce a simulation into an AI/ML network to allow for training of its semantically meaningful parameters. This whitepaper is a contribution to the Computational Frontier of Snowmass21.