[14:30] Coast-to-Coast Seminar: Classification in Genetic Programming: a Cooperative-Competitive Coevolution Approach

Speaker: Dr. Andrew McIntyre, Faculty of Computer Science, Dalhousie University

The method of Pareto dominance is increasingly being used within the context of coevolutionary approaches to Genetic Programming (GP). GP is a machine learning approach based on a neo-Darwinian metaphor for resolving the credit assignment problem. Coevolution provides a mechanism for establishing engagement between learner and domain; or resolving interactions between models with different behavioral contributions, thus problem decomposition. Pareto dominance has come to the fore as a formal mechanism for aiding both of these coevolutionary endeavors. In this presentation we will detail an approach to model building for the classification domain such that the Pareto coevolutionary scheme facilitates scalability to large data sets and acts as a natural mechanism for problem decomposition among cooperating classifiers. Specific comparisons will be made with classical machine learning algorithms and other GP classifiers.

If you are planning to attend this seminar, please contact a SHARCNET staff member at your local institution as AG rooms will only be opened for these seminars where someone has indicated they plan to attend.