NEMO: neural enhancement for multiobjective optimization
Abstract
In this paper, a neural network approach is presented to expand the Pareto-optimal front for multiobjective optimization problems. The network is trained using results obtained from the nondominated sorting genetic algorithm (NSGA-II) on a set of well-known benchmark multiobjective problems. Its performance is evaluated against NSGA-II, and the neural network is shown to perform extremely well. Using the same number of function evaluations, the neural network produces many times more non-dominated solutions than NSGA-II.
Authors: Aaron Garrett, Gerry Vernon Dozier, Kalyanmoy Deb
Published in: IEEE Congress on Evolutionary Computation (CEC) (2007)