知识卡片
固定点不保证全局最优
在相应正性和可微等条件下,BP固定点可与Bethe自由能的受约束驻点联系。不能把每个固定点一概称作全局最小值,也不能据消息不再变化断言得到真实边缘概率。优化近似目标与精确概率推断是不同层次的结论。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 概率图模型
原书出处 · 1 处
- · 查看原书依据
One can show that, in networks with loops, fixed points of the belief propagation algorithm (message configurations where the messages don't change with a message update) correspond to minima of a well-known approximation from the statistical physics community known as the Bethe free energy [513]. In practice, the solutions found by the loopy belief propagation algorithm can be quite good [346].
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:摘录中关于固定点不保证全局最优的定义、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:One can show that, in networks with loops, fixed points of the belief propagation algorithm (message configurations where the messages don't change with a message update) correspond to minima of a well-known approximation from the statistical physics community known as the Bethe free energy [513]. In practice, the solutions found by the loopy belief propagation algorithm can be quite good [346].