Stabilizing GANs' Training with Brownian Motion Controller
This addresses the critical problem of training instability for GAN users, offering a novel control-theoretic solution that is incremental in applying noise-based methods to enhance existing GAN frameworks.
The paper tackles the instability and lack of global convergence in GAN training by proposing a Brownian Motion Controller (BMC) based on control theory, which stabilizes training under StyleGANv2-ada frameworks with faster convergence, smaller oscillation, and improved FID scores.
The training process of generative adversarial networks (GANs) is unstable and does not converge globally. In this paper, we examine the stability of GANs from the perspective of control theory and propose a universal higher-order noise-based controller called Brownian Motion Controller (BMC). Starting with the prototypical case of Dirac-GANs, we design a BMC to retrieve precisely the same but reachable optimal equilibrium. We theoretically prove that the training process of DiracGANs-BMC is globally exponential stable and derive bounds on the rate of convergence. Then we extend our BMC to normal GANs and provide implementation instructions on GANs-BMC. Our experiments show that our GANs-BMC effectively stabilizes GANs' training under StyleGANv2-ada frameworks with a faster rate of convergence, a smaller range of oscillation, and better performance in terms of FID score.