Realizing GANs via a Tunable Loss Function

通过可调分类损失解决GAN训练不稳定问题

Addressing GAN Training Instabilities via Tunable Classification Losses

本文通过可调CPE损失重新表述GAN目标，提出(α_D,α_G)-GAN双目标框架，将训练稳定问题转化为f-散度最小化，并在多个数据集上验证其缓解训练不稳定的效果。

广义双判别器 GAN

Generalized Dual Discriminator GANs

本文提出 D2 α-GAN 并将其推广为广义双判别器 GAN，证明其极小极大目标等价于 f-散度与反向 f-散度组合的最小化，并通过二维合成实验验证优势。

$(α_D,α_G)$-GANs：通过双目标解决GAN训练不稳定性问题

$(α_D,α_G)$-GANs: Addressing GAN Training Instabilities via Dual Objectives

本研究提出双目标$(α_D,α_G)$-GANs框架，通过为生成器和判别器设置基于$α$-损失的不同目标，在理论上证明其可简化为$f$-散度最小化并推导估计误差上界，实验验证其能有效缓解GAN训练不稳定性。

Extracting Semantic Knowledge from GANs with Unsupervised Learning

Recently, unsupervised learning has made impressive progress on various tasks. Despite the dominance of discriminative models, increasing attention is drawn to representations learned by generative models and in particular, Generative Adversarial Networks (GANs). Previous works on the interpretation of GANs reveal that GANs encode semantics in feature maps in a linearly separable form. In this work, we further find that GAN's features can be well clustered with the linear separability assumption. We propose a novel clustering algorithm, named KLiSH, which leverages the linear separability to c…

最大切片Wasserstein距离及其在GANs中的应用

Max-Sliced Wasserstein Distance and its use for GANs

针对高维分布建模中GAN训练受距离度量样本复杂度影响的问题，分析切片Wasserstein距离的投影复杂度，提出兼具优异样本复杂度与更低投影复杂度的最大切片Wasserstein距离，可高效训练256x256分辨率的高维图像GAN。

基于一维条件GAN的（非）配对信号到信号翻译

(Un)paired signal-to-signal translation with 1D conditional GANs

本研究提出基于1D条件GAN的非配对sig2sig翻译方法，借鉴WaveGAN改造简化版CycleGAN，无需大幅改动原有架构即可实现1D信号翻译，在小型数据集上验证了含噪信号的域转换效果并量化差异。