Imagine a Binary Erasure Channel as depicted on Wikipedia.

One equation describing the mutual information is:

$$ \begin{align*}I(x;y) &= H(x) - H(x|y) \\ &= H(x) - p(y=0) \cdot 0 - p(y=?) \cdot H(x) -p(y=1)\cdot 0\end{align*}$$

Why is it "$p(y=?) \cdot H(x)$" and not "$p(y=?) \cdot H(x|y=?)$"?


If $y = ?$ then $H(x) = H(x|y=?)$ since $y = ?$ gives you no information about $x$.

  • $\begingroup$ I see, is that because p(x=0|y=?)=p(x=1|y=?) in a BEC? $\endgroup$ – user38931 May 6 '14 at 19:32
  • 1
    $\begingroup$ @user38931 The identity $H(x|y=?) = H(x)$ would be true even if $x$ were skewed. What we need is $p(x=0|y=?) = p(x=0)$. $\endgroup$ – Yuval Filmus May 6 '14 at 19:44

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.