Some challenges of calibrating differentiable agent-based models

Arnau Quera-Bofarull, Joel Dyer, Ani Calinescu, Michael Wooldridge

Keywords: differentiable agent-based model, Bayesian calibration, generalised variational inference, automatic differentiation

TL;DR: We identify some challenges on the application of vanilla automatic differentiation to agent-based models.

Abstract: Agent-based models (ABMs) are a promising approach to modelling and reasoning about complex systems, yet their application in practice is impeded by their complexity, discrete nature, and the difficulty of performing parameter inference and optimisation tasks. This in turn has sparked interest in the construction of differentiable ABMs as a strategy for combatting these difficulties, yet a number of challenges remain. In this paper, we discuss and present experiments that highlight some of these challenges, along with potential solutions.

Submission Number: 39