人工智能一种现代方法 - 随时间变化的概率推理

Artificial Intelligence A Modern Approach - Probabilistic Reasoning over Time

我目前正在阅读 Peter Norvig 的人工智能现代方法第 15 章随时间变化的概率推理,但我无法遵循过滤和预测的推导(第 572 页)。

Given the result of filtering up to time t, the agent needs to compute the result for t + 1 from the new evidence et+1,

P(Xt+1|e1:t+1) = f(et+1, P(Xt|e1:t)),

for some function f. This is called recursive estimation. We can view the calculation as being composed of two parts: first, the current state distribution is projected forward from t to t + 1; then it is updated using the new evidence et+1. This two-part process emerges quite simply when the formula is rearranged:

P(Xt+1|e1:t+1) = P(Xt+1|e1:t, et+1)     (dividing up the evidence)
= α P(et+1|Xt+1, e1:t) P(Xt+1|e1:t)     (using Bayes' rule)

在上一个论坛中使用贝叶斯法则结果如何?不应该

αP(e1:t,et+1|Xt+1) P(Xt+1)

它们都是正确的,除了两者之间的α不相同。诀窍在于,在 Norvig 对贝叶斯的应用中,e1:t 的条件始终保持不变。试想一下,它一开始就不存在。你仍然会拥有所有的身份。然后将该条件应用于等式的每一部分,所有恒等式都成立。

或者,也可以在不明确使用贝叶斯的情况下进行推导,而是仅使用联合概率和条件概率的定义(即 P(A|B)P(B)=P(A,B )=P(B|A)P(A)).

P(Xt+1|e1:t, et+1) P(et+1,e1:t) = P(e1:t, et+1|Xt+1) P(Xt+1)

扩大双方的联合概率得到

P(Xt+1|e1:t, et+1) P(et+1|e1:t) P(e1:t ) = P(et+1|Xt+1,e1:t) P (e1:t|Xt+1) P(Xt+1)

RHS 的最后两项可以重写为

P(Xt+1|e1:t, et+1) P(et+1|e1:t) P(e1:t ) = P(et+1|Xt+1,e1:t) P (Xt+1|e1:t) P(e1:t)

现在 P(e1:t) 取消(假设非零),我们有

P(Xt+1|e1:t, et+1) = α P(et+1|Xt+1,e1:t ) P(Xt+1|e1:t)

其中 α = 1 / P(et+1|e1:t).