28th Annual Student Research Conference
Swarm Intelligence for Multi-Objective Optimization
Son M. Nguyen
Dr. John Seiffertt, Faculty Mentor
Computational optimization is an important area in mathematics and the sciences. We investigate a relatively new approach, Particle Swarm Optimization (PSO), and modify it using a neural network to solve challenging optimization problems. PSO is a stochastic evolutionary algorithm that finds solutions on a very large scale search space with few assumptions. There is still much to be understood about how PSO works, and the results of our research contribute to this effort. Our study implements a neural network trained using PSO instead of traditional backpropagation methods. Within this framework, applications to optimizations with many goals are considered. Our new algorithms solve these multi-objective optimization problems more successfully than competing approaches. We assess different forms of PSO to identify which lead to superior performance. Our results show, visually and statistically, that our new methods achieve good performance on the tasks.
Keywords: Artificial intelligence, Particle swarm optimization, Evolutionary algorithm
Topic(s):Computer Science
Presentation Type: Oral Paper
Session: 205-4
Location: MG 1096
Time: 11:45