Apart from Mr. Melnik, author`s team of the article included Denis Nasonov, professor of Chair of High Performance Computing, Nikolay Butakov, PhD student, and Natalia Shindyapina, Master student of ITMO University.
Being a part of International Joint Conference on Computational Intelligence (IJCCI) ECTA brings together international researchers, engineers and practitioners on the areas of fuzzy computation, evolutionary computation and neural computation. The research project presented by Mikhail Melnik is a part of his Master`s thesis. Currently following a double degree program Mr. Melnik spends the third semester at University of Amsterdam.
The research focuses on optimization techniques for Workflow scheduling management system, which aims to process multiple data. The workflow technology also manages various business processes consisted of various targets.
“Workflow scheduling is very useful to researchers. Being involved in technological progress modern scientists deal with lots of data that are very difficult to process. For instance, optimizing and analyzing of information developed by Hadron Collider requires effective processing technology. A workflow scheduling system divides workload between computers. Finding new optimization algorithms is a great challenge for modern science. Moreover, dealing with information organization experts meet with such complicated problem as cloud environment…” said Mr. Melnik.
Most of relevant optimization systems are based on heuristic and metaheuristic algorithms. The last ones include evolutionary method, which contains coevolutionary algorithms. Their main feature is that they give an opportunity to provide several evolutionary processes at the same time making them dependent on each other.
According to Mikhail Melnik, the research focuses on three coevolutionary algorithms: coevolutionary genetic algorithm, coevolutionary particle swarm optimization and coevolutionary gravitation search algorithms. ITMO University researchers suppose that these well-known algorithms can be applied for workflow scheduling optimization.
Thanks to coevolutionary approach complicated tasks can be divided into simple challenges. Thus it is easier to optimize them. Furthermore these algorithms are more effective than other heuristic algorithms.
“The results of our research can be applied for effective over elaborate calculations. For example, currently researchers of High Performance Computing Chair develop a software suite for overflow prediction. Using coevolutionary algorithms one can determine when a dam should be closed. It`s very important because if it is closed too late some territories will be flooded. The premature closing leads to economic costs. That is why it is necessary to analyze a meteorological forecast quickly and correctly so as to make the right prediction,” said Ms. Shindyapina.