DEGAS: differentiable efficient generator search

Network architecture search achieves state-of-the-art results in various tasks such as classification and semantic segmentation. Recently, a reinforcement learning-based approach has been proposed for generative adversarial networks (GANs) search. In this work, we propose an alternative strategy GAN by using proxy task instead of common training. Our method is called differentiable efficient generator search, which focuses on efficiently finding the GAN. algorithm inspired differential global latent optimization procedure. This leads to both stable After found, it can be plugged into any existing framework For consistency-term GAN, use new model outperforms original inception score 0.25 CIFAR-10.