Understand loss function of GAN 理解GAN的损失函数

最新推荐文章于 2024-04-20 13:08:20 发布

davefighting

最新推荐文章于 2024-04-20 13:08:20 发布

阅读量2.1k

点赞数 2

分类专栏：技术类文章标签： GAN 损失函数 loss

本文链接：https://blog.csdn.net/zhouxiaowei1120/article/details/88533168

版权

技术类专栏收录该内容

16 篇文章 1 订阅

订阅专栏

注: 本文是作者的自我总结,主要作为个人总结记录, 欢迎大家批评,交流. https://zhouxiaowei1120.github.io/#blogs

Loss Function of GAN

$\min \limits_{G} \max \limits_D V(D, G) = \mathbb{E}_{x\sim p_{data}(x)}[\log D(x)] + \mathbb{E}_{z\sim p_{z}(z)}[\log (1 - D(G(z)))]$

My Understanding

$D$ try to maximize $V (D, G)$ , while $G$ try to minimize $V (D, G)$ . Larger the $V (D, G)$ , larger the $D$ , $D$ is better to discriminate the input. When we use cross entropy as loss function, because there is a minus between cross entropy and $V (D, G)$ , we should try to minimize cross entropy, which is equal to maximize $V (D, G)$ .

When it comes to $G$ , smaller the $V (D, G)$ , larger the $D (G (z))$ . That means $D$ try to classify $G (z)$ as true, which is just the goal of $G$ . When we use cross entropy as loss function, because there is a minus between cross entropy and $\mathbb{E}_{z\sim p_{z}(z)}[\log (1 - D(G(z)))]$ , we should try to minimize minus cross entropy.