compute H(X|Y) ( conditional probablity)

140 Views Asked by At

Can someone help me on this?

X = {$X_1,X_2,X_3,X_4$}

Y = {$Y_1,Y_2,Y_3,Y_4$}

Suppose p($X_i$) = p($Y_j$) = 1/4 (each X and each Y equally likely)

$1 \leq i, 4 \ge j$

and now suppose

$Y_1 : X_1 or X_2$

$Y_2 : X_3 or X_4$

$Y_3 : X_2 or X_3$

$Y_4 : X_1 or X_4$

what is $H(X|Y)$?


I tried to create a table, like $$ \left[ \begin{array}{ccc} & y1 & y2 & y3 & y4 \\ x1 & & 0 & 0 & \\ x2 & & 0 & & 0 \\ x3 & 0 & & & 0 \\ x4 & 0 & & 0 & \end{array} \right] $$ I know p(X) = 1/2, because only two choices. And p(Y)=1/4 is same.

Thank you


$\sum_{y}\sum_{x}p(y)p(x|y)log_2(p(x|y))$

$(1/4)p log_2 p + (1/4)((1/4)-p) log_2 ((1/4)-p)$

$4* [(1/4)p log_2 p + (1/4)((1/4)-p) log_2 ((1/4)-p)]$

$p log_2 p + ((1/4)-p) log_2 ((1/4)-p)$

1

There are 1 best solutions below

3
On

Let $p\mathop{:=}\mathsf P(X_1,Y_1)$.   Then since each column and each row must total $\tfrac 14$, we complete your table as follows:

$$\boxed{\begin{array}{|c|c:c:c|}\hline & Y_1 & Y_2 & Y_3 & Y_4 \\ \hline X_1 & p & 0 & 0 & \tfrac 14-p \\ \hdashline X_2 & \tfrac 14-p & 0 & p & 0 \\ \hdashline X_3 & 0 & p & \tfrac 14-p & 0 \\ \hdashline X_4 & 0 & \tfrac 14-p & 0 & p \\ \hline \end{array}}$$

The information you provided is insufficient to determine what $p$ actually is.

If the choice between each of the two $X_k$ for each $Y_h$ were unbiased, then $p=\tfrac 18$.   However, nothing indicates this is so.

But since there are the same two conditional probabilities for the two $X_k$ any given $Y_h$ (although which event they match differ), we can calculate the conditional entropy as:

$$\begin{align}\mathsf H(X\mid Y_h) & = - \sum_{k=1}^4 \mathsf P(X_k\mid Y_h)\log_2(\mathsf P(X_k\mid Y_h)) & \big[h\in\{1,2,3,4\}\big] \\ & = -\big( 4p \log_2 (4p) + 4(1-p)\log_2(4(1-p)) + 0\log 0 + 0\log 0\big) \\ & = -4\big( p \log_2(p) + (1-p) \log_2(1-p)+2 ) \big) \end{align}$$