Generative adversarial optimization

Abstract

Inspired by the adversarial learning in generative adversarial network, a novel optimization framework named Generative Adversarial Optimization (GAO) is proposed in this paper. This GAO framework sets up generative models to generate candidate solutions via an adversarial process, in which two models are trained alternatively and simultaneously, i.e., a generative model for generating candidate solutions and a discriminative model for estimating the probability that a generated solution is better than a current solution. The training procedure of the generative model is to maximize the probability of the discriminative model. Specifically, the generative model and the discriminative model are in this paper implemented by multi-layer perceptrons that can be trained by the back-propagation approach. As of an implementation of the proposed GAO, for the purpose of increasing the diversity of generated solutions, a guiding vector ever introduced in guided fireworks algorithm (GFWA) has been employed here to help constructing generated solutions for the generative model. Experiments on CEC2013 benchmark suite show that the proposed GAO framework achieves better than the state-of-art performance on multi-modal functions.