INDIVIDUAL-BASED MODELS: POWERFUL OR POWER STRUGGLE?

摘要: Individual-based models (IBMs) offer endless possibilities to explore various research questions but come with high model complexity and computational burden. Large-scale IBMs have become feasible the novel hardware architectures require adapted software. The increased also requires systematic exploration gain thorough system understanding. We elaborate on development of for vaccine-preventable infectious diseases active learning. Investment in IBM simulator code can lead significant runtime reductions. found large performance differences due data locality. Sorting population once, reduced simulation time by a factor two. Storing person attributes separately instead using objects seemed more efficient. Next, we improved up 70% structuring potential contacts based health status before processing disease transmission. learning approach present is iterative surrogate modelling model-guided experimentation. Symbolic regression used nonlinear response surface automatic feature selection. illustrate our an influenza vaccination. After optimizing parameter spade, observed inverse relationship between vaccination coverage clinical attack rate reinforced herd immunity. These insights be focus optimise activities, reduce both dimensionality decision uncertainty.