Life Sciences and Health Seminar, Joe Harrison, Leen Stougie

How GOMEA can be applied to Cartesian Genetic Programming

Join Zoom Meeting
https://cwi-nl.zoom.us/j/81535249101?pwd=ZDh3Y1kyZmxBUml1Tm00Rkx0NXBwdz09

Meeting ID: 815 3524 9101
Passcode: 967201

Title:      How GOMEA can be applied to Cartesian Genetic Programming
Speaker:    Joe Harrison
Abstract:   Genetic Programming (GP) has been a popular technique for symbolic regression. In GP expression trees are evolved by means of subtree crossover and point mutation to fit a function. In Cartesian Genetic Programming (CGP) a feed-forward acyclic graph is used rather than a tree and typically the only variation operators are point mutations. An advantage of using CGP is that nodes in the graph can be reused and more complex configurations such as graphs with skip-layers or multiple outputs are possible.

The Gene-Optimal Mixing Evolutionary Algorithm (GOMEA) aims to find linkage amongst components (e.g. nodes in a graph). GOMEA was successfully applied to GP. It groups tree nodes with linkage into subsets of nodes and exchanges these subsets as means of variation. In this presentation I will discuss how GOMEA can be applied to Cartesian Genetic Programming, the pros and cons of this application and how this affects node reuse.