Some challenges of calibrating differentiable agent-based models

Agent-based models (ABMs) are a promising approach Despite recent progress, the challenges involved in building to modelling and reasoning about complex and benefitting from differentiable ABMs remain underexplored, systems, yet their application in practice is impeded and there exists little guidance to practitioners by their complexity, discrete nature, and the interested in implementing and exploiting differentiable difficulty of performing parameter inference and ABMs. The aim of this paper is therefore to discuss some optimisation tasks. This in turn has sparked interest central challenges in applying AD to ABMs. in the construction of differentiable ABMs as a strategy for combatting these difficulties, yet