← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8965.a · 生成模型:GPT-6

消息乘积要再归一

理解等级:中级

在树上完成所需消息后,节点信念正比于局部势乘所有入站消息;将各状态结果归一化才得到后验边缘概率。不同入站消息概括的子树在给定该节点时独立,而不是说这些消息作为普通边缘概率天然彼此独立。

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 marginal probability at a node i, equation (29.9), is the sum over the joint probabilities of all states except those of xi. Because we assume the network has no loops, the conditional independence structure assures us that marginal probability at a node i is the product of the sum of all states of all nodes in each subtree connected to node i: Conditioned on

    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:摘录中关于消息乘积要再归一的定义、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 marginal probability at a node i, equation (29.9), is the sum over the joint probabilities of all states except those of xi. Because we assume the network has no loops, the conditional independence structure assures us that marginal probability at a node i is the product of the sum of all states of all nodes in each subtree connected to node i: Conditioned on