Memory-Guided Trust-Region Bayesian Optimization (MG-TuRBO) for High Dimensions

arXiv cs.LG / 4/13/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper addresses digital-twin calibration for traffic simulation as a costly, noisy, nonconvex optimization problem where each simulation trial is expensive and budgets are limited.
It compares genetic algorithms (GA) with several Bayesian optimization variants—classical BO, TuRBO, Multi-TuRBO, and a proposed Memory-Guided TuRBO (MG-TuRBO)—across real calibration tasks with 14 and 84 decision variables.
In the 14D setting, Bayesian optimization methods reach good calibration targets faster than GA, and MG-TuRBO performs comparably to the best BOM baselines.
In the 84D setting, MG-TuRBO shows noticeable advantages, especially when paired with an adaptive acquisition strategy.
The authors evaluate methods using final calibration quality, convergence behavior, and run-to-run consistency, concluding MG-TuRBO is particularly beneficial for high-dimensional traffic calibration and may generalize to other high-D problems.

Abstract

Traffic simulation and digital-twin calibration is a challenging optimization problem with a limited simulation budget. Each trial requires an expensive simulation run, and the relationship between calibration inputs and model error is often nonconvex, and noisy. The problem becomes more difficult as the number of calibration parameters increases. We compare a commonly used automatic calibration method, a genetic algorithm (GA), with Bayesian optimization methods (BOMs): classical Bayesian optimization (BO), Trust-Region BO (TuRBO), Multi-TuRBO, and a proposed Memory-Guided TuRBO (MG-TuRBO) method. We compare performance on 2 real-world traffic simulation calibration problems with 14 and 84 decision variables, representing lower- and higher-dimensional (14D and 84D) settings. For BOMs, we study two acquisition strategies, Thompson sampling and a novel adaptive strategy. We evaluate performance using final calibration quality, convergence behavior, and consistency across runs. The results show that BOMs reach good calibration targets much faster than GA in the lower-D problem. MG-TuRBO performs comparably in our 14D setting, it demonstrates noticeable advantages in the 84D problem, particularly when paired with our adaptive strategy. Our results suggest that MG-TuRBO is especially useful for high-D traffic simulation calibration and potentially for high-D problems in general.