USING GENETIC PROGRAMMING TO SYNTHESIZE MONOTONIC STOCHASTIC PROCESSES†

The automatic synthesis of stochastic concurrent processes is investigated. We use genetic programming to automatically evolve a set of stochastic π-calculus expressions that generate execution behaviour conforming to some supplied target behaviour. We model the stochastic π-calculus in a grammatically-guided genetic programming system, and machine model by Phillips and Cardelli. The behaviours of target systems are modelled as streams of numerical time series for different variables of interest. We were able to successfully evolve stochastic π-calculus systems that exhibited the target behaviors. Successful experiments considered target processes with continuous monotonic behaviours