Skip to content. | Skip to navigation

Personal tools

Tobias Helms, Stefan Rybacki, Roland Ewald, and Adelinde M Uhrmacher (2012)

An Adaptive Simulator for ML-Rules

Poster, , Winter Simulation Conference 2012.

Even the most carefully configured simulation algorithm may perform badly unless its configuration is adapted to the dynamics of the model. To overcome this problem, we apply methods from reinforcement learning to continuously re-configure an ML-Rules simulator at runtime. ML-Rules is a rule-based modeling language primarily targeted at multi-level microbiological systems. Our results show that, for models with sufficiently diverse dynamics, an adaptation of the simulator configuration may even outperform the best-performing non-adaptive configuration (which is typically unknown anyhow).