Versatile Auxiliary Classifier with Generative Adversarial Network (VAC+GAN)
This work addresses the problem of generating labeled adversarial samples for AI researchers, presenting an incremental improvement over existing conditional GAN methods.
The paper tackles the challenge of training conditional generators to produce labeled adversarial samples from specific distributions by introducing a framework that places a classifier parallel to the discriminator and backpropagates classification error through the generator. The method is versatile for any GAN variation and achieves superior results compared to similar approaches.
One of the most interesting challenges in Artificial Intelligence is to train conditional generators which are able to provide labeled adversarial samples drawn from a specific distribution. In this work, a new framework is presented to train a deep conditional generator by placing a classifier in parallel with the discriminator and back propagate the classification error through the generator network. The method is versatile and is applicable to any variations of Generative Adversarial Network (GAN) implementation, and also gives superior results compared to similar methods.