Towards a Better Global Loss Landscape of GANs
This addresses training challenges in generative adversarial networks, offering theoretical insights for improved stability, though it is incremental in advancing landscape analysis.
The paper tackles the problem of GAN training instability by analyzing the global loss landscape, proving that separable-GANs have many bad basins leading to mode collapse, while relativistic pairing GANs have no bad basins, with experiments showing RpGAN performs better on synthetic data.
Understanding of GAN training is still very limited. One major challenge is its non-convex-non-concave min-max objective, which may lead to sub-optimal local minima. In this work, we perform a global landscape analysis of the empirical loss of GANs. We prove that a class of separable-GAN, including the original JS-GAN, has exponentially many bad basins which are perceived as mode-collapse. We also study the relativistic pairing GAN (RpGAN) loss which couples the generated samples and the true samples. We prove that RpGAN has no bad basins. Experiments on synthetic data show that the predicted bad basin can indeed appear in training. We also perform experiments to support our theory that RpGAN has a better landscape than separable-GAN. For instance, we empirically show that RpGAN performs better than separable-GAN with relatively narrow neural nets. The code is available at https://github.com/AilsaF/RS-GAN.