Designing GANs: A Likelihood Ratio Approach
This work addresses the problem of designing and training GANs more reliably for researchers and practitioners, but it appears incremental as it builds on existing adversarial optimization frameworks.
The authors tackled the design of generative adversarial networks (GANs) by proposing a methodology for constructing consistent adversarial optimization problems and introducing a new likelihood ratio metric for monitoring convergence and stability during training. They demonstrated their approach by comparing various possibilities on well-known datasets using different neural network configurations.
We are interested in the design of generative networks. The training of these mathematical structures is mostly performed with the help of adversarial (min-max) optimization problems. We propose a simple methodology for constructing such problems assuring, at the same time, consistency of the corresponding solution. We give characteristic examples developed by our method, some of which can be recognized from other applications, and some are introduced here for the first time. We present a new metric, the likelihood ratio, that can be employed online to examine the convergence and stability during the training of different Generative Adversarial Networks (GANs). Finally, we compare various possibilities by applying them to well-known datasets using neural networks of different configurations and sizes.