To understand the coherence of Bayes' rule, imagine you are receiving pieces of evidence over time and you want to update your beliefs or probabilities based on this new evidence. Bayes' rule states that you can update your probabilities with each new piece of evidence as it comes in (sequentially), or you can wait and update your probabilities after you've gathered all the evidence (simultaneously). No matter which method you choose, the result will be the same. This is because Bayes' rule takes into account the prior probability and updates it in a consistent way, whether the evidence is considered all at once or in sequence.
To illustrate this, the text gives an example of conducting an experiment over a week. You could:
- Update your beliefs every day as new data comes in (sequential update).
- Wait until the end of the week and use all the collected data to update your beliefs once (simultaneous update).
Regardless of the approach, the final probability you calculate will be the same. This is what's meant by the rule being coherent—it consistently leads to the same result regardless of the updating method.
So, in essence, coherence in the context of Bayes' rule means that the order in which evidence is taken into account does not affect the final outcome of the probability calculation.
举个例子,假设我们在进行为期一周的实验,每天都会得到一些数据。我们可以每天使用贝叶斯法则根据当天的数据更新概率。或者我们可以去度假一周,等到周五下午回来时,用整周的数据一次性更新概率。无论哪种方法,最终得到的结果应该是相同的。
这就意味着,无论你是逐步更新你的概率还是等到最后一次性更新,贝叶斯法则都能保证你得出的概率是一致的。这就是“连贯性”的含义。