Machine learning aids agent-based model calibration, cutting time and error

Researchers have developed a machine learning-assisted approach to calibrate agent-based models (ABMs), which are often difficult to tune due to complex simulation landscapes. This new method embeds machine learning surrogates within genetic algorithms (GA) and particle swarm optimization (PSO) to screen candidate parameters, significantly reducing the need for costly black-box simulations. Evaluations on two distinct ABMs demonstrated that the best ML-assisted configurations achieved substantial reductions in root mean squared error (RMSE) and computation time compared to traditional optimization methods. The study also highlighted that the optimal combination of surrogate and optimization technique varies depending on the complexity of the ABM being calibrated. AI

IMPACT This research offers a more efficient method for tuning complex simulations, potentially accelerating scientific discovery in fields that rely on agent-based modeling.

RANK_REASON This is a research paper detailing a new methodology for calibrating agent-based models using machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• agent-based model
• Brock-Hommes asset-pricing model
• Duguma Yeshitla Habtemariam
• genetic algorithm
• Island growth model
• particle swarm optimization

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →