← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8967.d · 生成模型:GPT-6

逐点最优不等于联合最优

理解等级:中级

分别取各节点边缘分布的最大状态,优化的是相应逐点决策,未必组成联合后验概率最大的配置。联合MAP需要考虑状态组合的相容性,树上可用最大积和回溯;一般有环图则需要适合目标的近似或专门优化算法。

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

      Instead of summing over the states of other nodes, we are sometimes interested in finding the x that maximizes the joint probability. The argmax operator passes through constant variables just as the summation sign did. This leads to an alternate version of the BP algorithm, equation (29.14), with the summation (of multiplying the vector message products by the node compatibility matrix) replaced with argmax. This

    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:摘录中关于逐点最优不等于联合最优的定义、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:Instead of summing over the states of other nodes, we are sometimes interested in finding the x that maximizes the joint probability. The argmax operator passes through constant variables just as the summation sign did. This leads to an alternate version of the BP algorithm, equation (29.14), with the summation (of multiplying the vector message products by the node compatibility matrix) replaced with argmax. This