Following on from the workshop we held at the Social Simulation Conference, we are inviting people to attend a two-day workshop in London to build on co-constructed visions for exascale computing supporting agent-based modelling. The abstract for the workshop is provided below.
Exascale computing is high-end computation entailing 10^18 (a billion-billion) floating-point operations per second (FLOPS). By comparison, a modern laptop is capable of roughly a billion FLOPS. The UK is planning to commission an exascale computer, and the Engineering and Physical Sciences Research Council has developed the ExCALIBUR programme to prepare the research community. As part of this, the Exascale Agent-Based Modelling for PoLicy Evaluation in Real-time (ExAMPLER) project is undertaking a gap analysis to investigate the needs of the empirical agent-based modelling community in using an exascale computing facility. In terms of raw numbers, the computer speeds entailed could transform agent-based modelling, as simulation experiments that take several days or weeks on existing high-performance computing facilities could be completed in under a second at exascale. However, typical-case agent-based modelling is still done on personal computers; cloud and cluster computing with agent-based models is the exception rather than the rule. This workshop will explore with participants their transformative long-term visions for agent-based modelling underpinned by exascale computing, along with associated software and institutional requirements. We focus on agent-based modelling applied in the social sciences, where collaborations with industry and policy, particularly in supporting decision-making in complex environments, are a common use-case for agent-based models.
