Abstract: Decision making for complex systems are based on multi-criterion-optimisation. For instance, a task assignment in a distributed computer system may reduce both the total cost of a program run and a workload of the bottleneck computer. Moreover, it can decrease the cost of computers if a selection of the computer sort is carried out. A total amount of the system performance is another measure that can be minimized by task distribution. The probability that all computers remain fault-free during the execution of the modules assigned to computers is the first criterion of evaluation task assignments.
Above benchmark problem can be formulated as a multiobjective combinatorial optimisation question, which is solved by an approach based on artificial intelligence. It is applied for finding the subset of Pareto-optimal solutions.
Genetic algorithms, evolutionary algorithms, evolution strategies and genetic programming are the alternative evolutionary approaches to the modern metaheuristic multicriteria optimisation methods such as simulated annealing, tabu search or Hopfield models of neural networks. Especially, genetic programming paradigm is implemented as a genetic algorithm written in the Matlab language. Chromosomes are generated as the Matlab functions and then genetic operators are applied for finding Pareto-suboptimal task assignment. Results are compared with outcomes obtained by an adaptive evolutionary algorithm.

Brief biography of the speaker:
Jerzy Balicki received the M.Sc. and Ph.D. degrees in Computer Science from University of Military Technology, Warsaw, Poland in 1982 and 1987, respectively. Then, he achieved habilitation D.Sc. from Technical University of Poznan in 2001. He was admitted as a university professor at Naval University of Gdynia in 2002. He is an author of three books and more than 120 scientific papers related to artificial intelligence, distributed computer systems, and decision support systems.