This thesis begins by examining current research in both genetic programming and particle swarm optimization. A hybrid is then proposed, merging particle swarm optimization with genetic programming. Feature sets are used to describe a LISP program. These sets map the LISP program onto an n-dimensional space wherein the particle swarm algorithm can function with minimal modifications. The modification made to the particle swarm algorithm replaces the velocity component of the algorithm with a new component. Three parameters are explored, one of which is a new parameter added to the particle swarm algorithm. Each parameter option is tested against each other as well as a standard genetic program. The test consists of four geometric functions. |