arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.09134cs.CVcs.LGeess.IV
收藏

AlphaGAN:生成对抗网络的全微分结构搜索

AlphaGAN: Fully Differentiable Architecture Search for Generative Adversarial Networks

Zhejiang University(浙江大学);JD Explore Academy(京东探索研究院);Tencent AI Lab(腾讯人工智能实验室);Tencent Data Platform(腾讯数据平台部);Microsoft(微软)

Yuesong Tian, Li Shen, Li Shen, Guinan Su, Zhifeng Li, Wei Liu

摘要:

生成对抗网络(GANs)被公式化为极大极小博弈问题,其中生成器试图通过对抗鉴别器的对抗学习来接近真实数据分布。问题固有的复杂性对提高生成网络的性能提出了挑战。在这项工作中,我们的目标是从网络架构的角度来促进模型学习,将自动架构搜索的最新进展纳入GANs。为此,我们为生成对抗网络提出了一个完全可区分的搜索框架,称为alphaGAN。搜索过程被形式化为求解一个双层minimax优化问题,其中外层目标是在发电机和内部网络参数的条件下寻求一个合适的网络结构,以达到纯纳什均衡,而内部网络参数是在传统的GAN损失下优化的。整个优化过程采用一阶优化方法,以完全可微的方式交替最小化两级目标,使体系结构搜索能够在巨大的搜索空间内完成。CIFAR-10和STL-10数据集上的大量实验表明,我们的算法可以获得高性能的架构,只有3-GPU小时,在一个单一的GPU在搜索空间组成的约2?1011种可能的配置。我们还提供了一个全面的分析搜索过程的行为和搜索架构的属性,这将有利于进一步研究生成模型的架构。预训练的模型和代码可在https://github.com/yuesongtian/AlphaGAN上获得。

英文摘要:

Generative Adversarial Networks (GANs) are formulated as minimax game problems, whereby generators attempt to approach real data distributions by virtue of adversarial learning against discriminators. The intrinsic problem complexity poses the challenge to enhance the performance of generative networks. In this work, we aim to boost model learning from the perspective of network architectures, by incorporating recent progress on automated architecture search into GANs. To this end, we propose a fully differentiable search framework for generative adversarial networks, dubbed alphaGAN. The searching process is formalized as solving a bi-level minimax optimization problem, in which the outer-level objective aims for seeking a suitable network architecture towards pure Nash Equilibrium conditioned on the generator and the discriminator network parameters optimized with a traditional GAN loss in the inner level. The entire optimization performs a first-order method by alternately minimizing the two-level objective in a fully differentiable manner, enabling architecture search to be completed in an enormous search space. Extensive experiments on CIFAR-10 and STL-10 datasets show that our algorithm can obtain high-performing architectures only with 3-GPU hours on a single GPU in the search space comprised of approximate 2 ? 1011 possible configurations. We also provide a comprehensive analysis on the behavior of the searching process and the properties of searched architectures, which would benefit further research on architectures for generative models. Pretrained models and codes are available at https://github.com/yuesongtian/AlphaGAN.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑