Overcoming Mode Collapse with Adaptive Multi Adversarial Training
This addresses the mode collapse issue in GANs for generative modeling applications, representing an incremental improvement by plugging into existing frameworks.
The paper tackled the mode collapse problem in Generative Adversarial Networks (GANs) by linking it to catastrophic forgetting in discriminators and introduced an adaptive multi-adversarial training method that spawns additional discriminators to remember previous modes, resulting in mitigation of mode collapse and improved standard GAN evaluation metrics on several datasets.
Generative Adversarial Networks (GANs) are a class of generative models used for various applications, but they have been known to suffer from the mode collapse problem, in which some modes of the target distribution are ignored by the generator. Investigative study using a new data generation procedure indicates that the mode collapse of the generator is driven by the discriminator's inability to maintain classification accuracy on previously seen samples, a phenomenon called Catastrophic Forgetting in continual learning. Motivated by this observation, we introduce a novel training procedure that adaptively spawns additional discriminators to remember previous modes of generation. On several datasets, we show that our training scheme can be plugged-in to existing GAN frameworks to mitigate mode collapse and improve standard metrics for GAN evaluation.