Parameter Optimization of a Cellular Automaton Model in Distributed Environment

Abstract

We present a distributed software environment to search for optimal parameters of the Cellular Potts Model using data from cell culture experiments. The software architecture follows a master-worker pattern. The worker nodes are executing simulations of the model with given parameter sets and send the computed objective function values back to the master. The master node implements an optimization method, a genetic algorithm in our case, and a task scheduler to distribute the generated simulation tasks among the workers through the network. Results showed that an optimal combination of parameters can be found for the model, and the efficiency scales linearly with the number of workers.