Can Genetic Programming Tackle More Complex Problems?
Abstract: Genetic Programming (GP) involves artificially evolving programs to solve problems. It has proved to be an effective machine learning technique in a wide range of domains from financial trading and bioinformatics to board games and even art. GP is typically highly productive during the initial search phases of evolution but it stagnates before deep levels of complexity are acquired. The aim of my research is to investigate whether genetic programming can tackle more complex problems through the long term acquisition of novel features. This may be at the expense of the speed of the initial search if necessary. Experiments which investigate extending the complexity of GP for long runs require substantial computational resources. Much of my work so far has focussed on harnessing Graphics Processing Unit (GPU) technology to meet this need. This is achieved by simultaneously using two graphics cards to evaluate GP individuals using nVidia’s “CUDA” technology. Brian Gannon Title: G
