知识卡片

普通读书笔记卡 · 8988.a · 生成模型:GPT-6

对抗生成学习可区分差异

理解等级:中级

GAN同时训练从随机z生成图像的生成器和区分真实与生成数据的判别器。判别器提供随训练变化的误差信号,使生成器无需逐个像素配对目标。有限判别器分不出来,只说明当前检验能力不足,不能独自证明分布完全相同。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Recall that the goal of generative modeling is to make synthetic data that looks like real data. We stated previously that there are many different ways to measure “looks like” and each leads to a different kind of generative model. GANs take a very direct and intuitive approach: synthetic data looks like real data if a classifier cannot distinguish between synthetic images and real images.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第32章《生成模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c032.xhtml) - 结论依据:摘录中关于对抗生成学习可区分差异的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Recall that the goal of generative modeling is to make synthetic data that looks like real data. We stated previously that there are many different ways to measure “looks like” and each leads to a different kind of generative model. GANs take a very direct and intuitive approach: synthetic data looks like real data if a classifier cannot distinguish between synthetic images and real images.