IEEE CEC 2025 – 4th Special Session on Partitioning, Clustering and Decomposition in Evolutionary Computation

Overview and Scope

Partitioning points out techniques in which the samples are divided into different groups in terms of some criteria. Clustering is a special type of partitioning approach in which the samples are categorised into a group according to their similarities. Partitioning can be applied for different purposes in evolutionary computation such as improving the performance, finding a promising region, solving a large-scale problem, and finding a proper value for a specific parameter. Generally speaking, partitioning and clustering can be seen in different levels of an algorithm. The most important ones are as listed follows:

• Partitioning and clustering in the population level: partitioning the whole or a subset of the population is more popular among others. Multi-population can outperform the performance since they work by partitioning the population
  into smaller subpopulations which can move towards different optima. As a result, preserving diversity and helping algorithms to escape from local optima is easier. Also, Multi-population can be used for other goals such
  as finding a promising region.
• Partitioning and clustering at the individual level: these strategies try to group variables. For example, Cooperative Coevolutionary algorithms identify dependent variables and then group them into subproblems.
•  Partitioning and clustering in the parameter level: these strategies work based on parameters and not population.  There is not much research in this category.

The goal of this special session is the use of partitioning and clustering approaches for evolutionary computation, covering all different evolutionary paradigms. The aim is to investigate both the new theories and methods in different evolutionary paradigms in different domains such as image processing, pattern recognition, scheduling, and transportation.