知识卡片
可生成不代表能直接计算密度
赋值Y=f(X,N)使采样容易实现,但当f任意、不可逆或维度改变时,观测Y的密度未必能直接求出。基于似然的学习还需要适当密度计算、潜变量积分或近似推断方法;普通Python赋值不会自动提供这些概率操作。
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
value (in more precise terms, the likelihood) of each value of Y in the training data. Those values are used in probabilistic inference algorithms like MCMC. But if I write a statement that represents an assignment function, like y = f(x, ny), Pyro doesn’t automatically know how to calculate probability values for Y, especially since as far as Pyro is concerned, f(.) can be anything.
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:摘录中关于可生成不代表能直接计算密度的定义、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:value (in more precise terms, the likelihood) of each value of Y in the training data. Those values are used in probabilistic inference algorithms like MCMC. But if I write a statement that represents an assignment function, like y = f(x, ny), Pyro doesn’t automatically know how to calculate probability values for Y, especially since as far as Pyro is concerned, f(.) can be anything.