Analyzing the Components of Distributed Coevolutionary GAN Training

08/03/2020
by   Jamal Toutouh, et al.
0

Distributed coevolutionary Generative Adversarial Network (GAN) training has empirically shown success in overcoming GAN training pathologies. This is mainly due to diversity maintenance in the populations of generators and discriminators during the training process. The method studied here coevolves sub-populations on each cell of a spatial grid organized into overlapping Moore neighborhoods. We investigate the impact on the performance of two algorithm components that influence the diversity during coevolution: the performance-based selection/replacement inside each sub-population and the communication through migration of solutions (networks) among overlapping neighborhoods. In experiments on MNIST dataset, we find that the combination of these two components provides the best generative models. In addition, migrating solutions without applying selection in the sub-populations achieves competitive results, while selection without communication between cells reduces performance.

READ FULL TEXT

page 9

page 10

research
02/10/2021

Signal Propagation in a Gradient-Based and Evolutionary Learning System

Generative adversarial networks (GANs) exhibit training pathologies that...
research
06/25/2021

Fostering Diversity in Spatial Evolutionary Generative Adversarial Networks

Generative adversary networks (GANs) suffer from training pathologies su...
research
04/07/2020

Data Dieting in GAN Training

We investigate training Generative Adversarial Networks, GANs, with less...
research
05/29/2019

Spatial Evolutionary Generative Adversarial Networks

Generative adversary networks (GANs) suffer from training pathologies su...
research
11/30/2018

Lipizzaner: A System That Scales Robust Generative Adversarial Network Training

GANs are difficult to train due to convergence pathologies such as mode ...
research
01/05/2018

Dynamic Island Model based on Spectral Clustering in Genetic Algorithm

How to maintain relative high diversity is important to avoid premature ...
research
05/22/2016

Evolutionary Demographic Algorithms

Most of the problems in genetic algorithms are very complex and demand a...

Please sign up or login with your details

Forgot password? Click here to reset