知识卡片

普通读书笔记卡 · 8980.b · 生成模型:GPT-6

生成噪声提供变化来源

理解等级:初级

生成模型中的噪声常指抽样得到的随机输入,用来选择产生哪一个样本;它不只是待清除的传感器干扰。扩散中的前向加噪与生成器潜在随机性又有不同角色。理解算法时先说明随机量来自哪里、在何处进入,以及怎样影响输出。

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

      The z-vector is also sometimes called noise, but noise has a very different meaning here than in signal processing. In signal processing, noise is a nuisance to get rid of. In generative modeling, the noise is the signal; it determines exactly which image to generate.

    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:摘录中关于生成噪声提供变化来源的定义、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:The z-vector is also sometimes called noise, but noise has a very different meaning here than in signal processing. In signal processing, noise is a nuisance to get rid of. In generative modeling, the noise is the signal; it determines exactly which image to generate.