Beyond Local Nash Equilibria for Adversarial Networks
This addresses the instability and quality issues in GAN training for generative modeling, offering a more robust theoretical framework, though it is an incremental improvement over existing methods.
The paper tackles the problem that GAN training often converges to local Nash equilibria (LNEs) with poor quality guarantees, by modeling GANs as finite games in mixed strategies to ensure every LNE is a Nash equilibrium (NE). It proposes a method proven to converge to resource-bounded NEs, empirically showing reduced mode collapse and less exploitable solutions compared to GANs and MGANs.
Save for some special cases, current training methods for Generative Adversarial Networks (GANs) are at best guaranteed to converge to a `local Nash equilibrium` (LNE). Such LNEs, however, can be arbitrarily far from an actual Nash equilibrium (NE), which implies that there are no guarantees on the quality of the found generator or classifier. This paper proposes to model GANs explicitly as finite games in mixed strategies, thereby ensuring that every LNE is an NE. With this formulation, we propose a solution method that is proven to monotonically converge to a resource-bounded Nash equilibrium (RB-NE): by increasing computational resources we can find better solutions. We empirically demonstrate that our method is less prone to typical GAN problems such as mode collapse, and produces solutions that are less exploitable than those produced by GANs and MGANs, and closely resemble theoretical predictions about NEs.