← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8965 · 生成模型:GPT-6

排除回信避免重复计入

理解等级:中级

计算j给i的消息时,不使用i刚发给j的消息。树上删除该边会分成两个子树,发送侧的信息不应包含接收侧已经提供的证据。这个排除规则是精确分解的一部分,不是任意丢弃一个输入的启发式。

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

      1. Multiply together all messages (independent probabilities) coming in to node j, except for the message from node i back to node j, 2. Multiply by the pairwise compatibility function ψij(xi, xj) (also independent of the other multiplicands). 3. Marginalize over the variable xj. These steps are summarized in this equation to compute the message from node j to node i: (29.14) where η(j)\i means

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第29章《概率图模型》(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/c029.xhtml) - 结论依据:摘录中关于排除回信避免重复计入的定义、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:1. Multiply together all messages (independent probabilities) coming in to node j, except for the message from node i back to node j, 2. Multiply by the pairwise compatibility function ψij(xi, xj) (also independent of the other multiplicands). 3. Marginalize over the variable xj. These steps are summarized in this equation to compute the message from node j to node i: (29.14) where η(j)\i means