接着，作者举例，如何将信息论与概率论挂钩起来 —— 经典的就是盒子模型，或者等价的，作者举例的urn模型：
we can consider information theory in the context of a generic random event: repeatedly drawing a marble at random from an urn containing blue, green, yellow, and gray marbles.
有放回的抽样，用这个来解释信息论：
a) A sequence of two marbles is drawnat random (with replacement) from an urn, giving rise to
b) a probability distribution over the16 possible two-color sequences.
c) Learning that a proposition about the two colors drawn istrue enables the elimination of certain outcomes. For example, learning neither marble is blueeliminates 7/16 possibilities containing 3/4of the the probability mass.
而转换到信息论的角度：
Eliminating probability mass, reducing uncertainty about the outcome, and gaining information are all mathematically equivalent.
如何度量信息（Information gained）呢？
Reduction of 50% of the probability mass corresponds to 1 bit of information....
Each time we can rule out half of the probability mass, we have gained exactly 1 bit of information.
对应的公式是：
链接完成。
进一步，对于每一个random event，就有：
Since we can calculate the information provided by each outcome, and we know each outcome’sprobability, we can compute the probability-weighted average amount of information provided by arandom event, otherwise known as the entropy.
信息熵的公式是：
一个有趣的结论是：
It might seem that, since lower probability outcomes contain more information, that randomevents containing many extremely rare outcomes will have the highest entropy. In fact, the oppositeis true: random events with probability spread as evenly as possible over their possible outcomeswill have the highest entropy.
网友评论