条件期望函数
设 和 是两个随机变量,对于任意一个实数 都仅存在一个实数 与之对应,因此可以构造随机变量 的映射 ,称 为条件期望函数 (CEF)。
Law of Iterated Expectations
Theorem
Proof
离散型随机变量
连续型随机变量
推论:将 整体视为一个随机变量
Best Predictor
Theorem Let be any function of , then
Proof
the first term
the last term
therefore, take
因此,条件期望函数 是给定 对 做出的最小平方预测。