Optimization of following model parameters in a computing cluster by asynchronous modification of the ant colony method

V. A. Sudakov, Yu. P. Titov, P. M. Ivanova, T. V. Sivakova


Abstract: The paper is devoted to the parallel modification of the ant colony method for the parametric optimization problem in order to conserve cluster computing resources. The approach of dividing the ant colony method into a stream of parallel pathfinding and a stream of sending parameter values to the model on a separate computer is considered. It is proposed to introduce a copy of the graph for continuous path search by agents without blocking the process of updating pheromones on the graph. For interaction with the computing cluster, an interaction protocol is proposed, with the ability to support multi-packet data sending.

Keywords: ant colony method, iteration of parameter values, parallel computing, multiextremeality, cluster.