← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8965.b · 生成模型:GPT-6

两趟遍历得到全部边缘

理解等级:中级

任选树根,先从叶向根计算消息,根即可汇集全树证据;再从根向叶计算反向消息,便可得到各节点边缘。每条边各方向所需消息被复用,避免为每个目标节点重新求和。同步调度也可行,但必须遵守依赖已就绪的条件。

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

      A node can send a message whenever all the incoming messages it needs have been computed. We can compute the outgoing messages from leaf nodes in the graphical model tree, since they have no incoming messages other than from the node to which they are sending a message, which doesn't enter in the outgoing message computation. Two natural message passing protocols are consistent with that

    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:摘录中关于两趟遍历得到全部边缘的定义、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:A node can send a message whenever all the incoming messages it needs have been computed. We can compute the outgoing messages from leaf nodes in the graphical model tree, since they have no incoming messages other than from the node to which they are sending a message, which doesn't enter in the outgoing message computation. Two natural message passing protocols are consistent with that