Agent-based model simulation costs cut with neural posterior estimation
Simulation-Efficient Double-Flow Neural Posterior Estimation for Agent-Based Models
Multiagent Systems
Summary
Simulating agent-based models to find parameter values can be very slow and expensive. The authors propose a new method called Double-Flow Neural Posterior Estimation (DF-NPE) that mixes real simulations with synthetic data generated by neural networks, making this process much faster. They tested their method on a financial model and showed it uses 90 percent fewer direct simulations while still providing better estimates and being faster overall. This approach helps reduce the high cost usually needed for such simulations.
What this means in practice
- •For financial modelers: Estimate parameters in asset-pricing simulations with fewer and faster runs by using the proposed neural posterior approach.
- •For epidemiological modelers: Reduce costly simulations when inferring parameters in disease spread models by augmenting real data with synthetic examples.
Authors
Junyao Zhang, Jinglai Li, Junqi Tang
Abstract
Neural posterior estimation (NPE) enables Bayesian inference of the posterior parameter in agent-based models (ABMs) with intractable likelihoods, but simulating training data can be computationally expensive. We propose Double-Flow Neural Posterior Estimation (DF-NPE), which combines learned time-series summaries with two conditional normalizing flows for synthetic-data generation and posterior estimation. The framework supplements direct simulations with synthetic training pairs and supports both pooled and ordered training strategies. Experiments in two settings of the Brock-Hommes asset-pricing model show that DF-NPE can reduce direct simulations by 90 percent while achieving lower mean posterior discrepancies and runtime than standard NPE. Performance varies with the allocation of directly simulated and synthetic data and the training strategy. These results establish learned summary augmentation as a practical approach to reducing simulation costs for Bayesian posterior inference of parameters in ABMs.