Skip to yearly menu bar Skip to main content

Workshop: Information-Theoretic Methods for Rigorous, Responsible, and Reliable Machine Learning (ITR3)

Realizing GANs via a Tunable Loss Function

Gowtham Raghunath Kurri · Tyler Sypherd · Lalitha Sankar

Abstract: We introduce a tunable GAN, called $\alpha$-GAN, parameterized by $\alpha \in (0,\infty]$, which interpolates between various $f$-GANs and Integral Probability Metric based GANs (under constrained discriminator set). We construct $\alpha$-GAN using a supervised loss function, namely, $\alpha$-loss, which is a tunable loss function capturing several canonical losses. We show that $\alpha$-GAN is intimately related to the Arimoto divergence, which was first proposed by \"{O}sterriecher (1996), and later studied by Liese and Vajda (2006). We posit that the holistic understanding that $\alpha$-GAN introduces will have practical benefits of addressing both the issues of vanishing gradients and mode collapse.

Chat is not available.