Generative Minimization Networks: Training GANs Without Competition
This addresses instability and convergence issues in GAN training for practitioners in generative modeling.
The paper tackles the convergence problems in GAN training by proposing a new objective that avoids the min-max structure, using a duality gap approach from game theory, and provides novel convergence guarantees showing it outperforms existing methods.
Many applications in machine learning can be framed as minimization problems and solved efficiently using gradient-based techniques. However, recent applications of generative models, particularly GANs, have triggered interest in solving min-max games for which standard optimization techniques are often not suitable. Among known problems experienced by practitioners is the lack of convergence guarantees or convergence to a non-optimum cycle. At the heart of these problems is the min-max structure of the GAN objective which creates non-trivial dependencies between the players. We propose to address this problem by optimizing a different objective that circumvents the min-max structure using the notion of duality gap from game theory. We provide novel convergence guarantees on this objective and demonstrate why the obtained limit point solves the problem better than known techniques.