SSCo 2013 - Special Session on Coevolution

Bio-Inspired methodologies that are based on the natural coevolutionary process have been applied successfully to solve a variety of machine learning problems. In particular, competitive coevolution is used to solve difficult adversarial problems such as games whereby the target functions are unknown and that training samples are unavailable for supervised learning methods. Competitive coevolution seeks to solve these problems naturally with one population consisting of candidate solutions (e.g. game strategies) and another population consisting of test cases (e.g. test strategies) that interact and undergo adaptation in a manner that promotes the search for problem solutions while using typically a small number of representative test cases that are discovered. Other research studies have been made in the framework of cooperative coevolution and its novel use to solve complex real-world learning problems that are amenable to divide-and-conquer approaches. Examples include ensemble learning for classification tasks and data mining through Bayesian networks. Furthermore, recent theoretical studies have been made for coevolutionary learning. These include quantitative performance analysis of coevolutionary algorithms through the generalization framework from machine learning, which provide the means for in-depth analysis how specific designs of components (e.g., selection and variation operators) can affect the performance of coevolutionary learning. This special session aims to bring together researchers in theoretical aspects and practitioners in the real-world problem solving applications of coevolution.
For more information, see the special session web site http://baggins.nottingham.edu.my/~khczcsy/ideal201....