Consider a binary message in which $0$ has has probability $1/3$ and $1$ has probability $2/3$. What value of $H$ should be assign?

I know that you split up $1$ into two messages $1a$ and $1b$. Then I think you have to use conditional probability here. Am I on the right track? So $$\left[\text{entropy of} \ (0,1a, 1b) \ \text{message} \right] = \left[ \text{entropy of} \ (0,1) \ \text{message} \right] + \text{something}$$

  • $\begingroup$ What does "split up 1 into two messages 1a and 1b" mean? $\endgroup$
    – Emre
    May 1, 2011 at 4:14
  • $\begingroup$ I agree with Emre. There is no reason to split 1 into two messages 1a and 1b (unless you are looking at a problem which explicitly involves such a split, in which case you should specify it properly in your question.) $\endgroup$
    – svenkatr
    May 1, 2011 at 6:12

1 Answer 1


Could you define $H$ so others don't have to look? Wikipedia says $H(x)=-\sum p(x_i)\log(p(x_i)).$ You have the $p(x_i)$ so what more do you need? Entropy is additive.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.