Unrolled Generative Adversarial Networks
Luke Metz Google Brain、Ben Poole Stanford University、David Pfau Google DeepMind、Jascha Sohl-Dickstein Google Brain Work done as a member of the Google Brain Residency program (g.co/brainresidency)Work completed as part of a Google Brain internship
Abstract
We introduce a method to stabilize Generative Adversarial Networks (GANs) by defining the generator objective with respect to an unrolled optimization of the discriminator. This allows training to be adjusted between using the optimal discriminator in the generator’s objective, which is ideal but infeasible in practice, and using the current value of the discriminator, which is often unstable and leads to poor solutions. We show how this technique solves the common problem of mode collapse, stabilizes training of GANs with complex recurrent generators, and increases diversity and coverage of the data distribution by the generator.
中文速览
训练生成对抗网络（GAN）时，生成器经常陷入"模式崩塌"（mode collapse）——只生成少数几种雷同样本，而非覆盖真实数据的多样分布，根本原因在于生成器每次更新时只看到当前判别器的反应，而非判别器最优状态下的反应。本文提出"展开优化"（unrolled optimization）方法：在更新生成器参数时，不直接使用当前判别器，而是将判别器的未来K步梯度更新过程"展开"并可微地嵌入生成器的目标函数，让生成器能够预判判别器会如何响应自己的动作，从而主动避免被轻易识破的崩塌行为。实验结果表明，该方法有效缓解了模式崩塌问题，稳定了复杂循环生成器的训练过程，并显著提升了生成分布对真实数据分布的覆盖度和多样性。这项工作为GAN训练不稳定这一长期痛点提供了一个有理论依据且实践可行的解决思路，仅需调整展开步数K即可在标准训练与理想训练目标之间灵活权衡。
原文 arXiv:1611.02163;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1611.02163v4