知识卡片
流用变量替换计算输出密度
对固定条件x,若Y=f_x(N)是可微可逆变换,则log p(Y|x)=log p_N(f_x⁻¹(Y))+log|det J_f_x⁻¹(Y)|。Jacobian修正体积伸缩,不能只把输出送入基础分布。可逆采样与显式密度使流适用于某些结构赋值建模。
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
- · 结构因果模型
原书出处 · 1 处
- · 查看原书依据
change-of-variable formula from probability theory). That automatic calculation relies on monotonicity; our causal “how” assumption is that the relationship between enzyme concentration and protein abundance is monotonic, and normalizing flows give us monotonicity.
参考来源
- 位置:《Causal AI》第06章《结构因果模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-6.xhtml)
- 结论依据:摘录中关于流用变量替换计算输出密度的定义、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:change-of-variable formula from probability theory). That automatic calculation relies on monotonicity; our causal “how” assumption is that the relationship between enzyme concentration and protein abundance is monotonic, and normalizing flows give us monotonicity.