← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8997.e · 生成模型:GPT-6

条件先验可依赖输入

理解等级:中级

一般条件潜变量模型写为p(y|x)=∫p(y|z,x)p(z|x)dz。书中进一步选择与x独立的先验p(z),这是方便的模型设定,不能当作所有cVAE的必要定义。独立先验也不强制训练后潜坐标只表示唯一的“剩余因素”。

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

      In conditional VAEs (cVAEs), we restrict our attention to latent variables z that are independent of the inputs we are conditioning on, so we have: (34.7) This corresponds to this graphical model: The idea is that z should only encode bits of information about y that are independent from whatever x already tells us about y. For example, suppose that we are trying to predict

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第34章《条件生成模型》(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/c034.xhtml) - 结论依据:摘录中关于条件先验可依赖输入的定义、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:In conditional VAEs (cVAEs), we restrict our attention to latent variables z that are independent of the inputs we are conditioning on, so we have: (34.7) This corresponds to this graphical model: The idea is that z should only encode bits of information about y that are independent from whatever x already tells us about y. For example, suppose that we are trying to predict