This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: ECE 1502 Information Theory Problem Set 5 solutions 1 April 11, 2006 9.1 Differential Entropy. (a) Exponential distribution. h ( f ) = Z e x [ln  x ] dx (1) = ln + 1 nats. (2) = log e bits. (3) (b) Laplace density. h ( f ) = Z  1 2 e  x  [ln 1 2 + ln   x  ] dx (4) = ln 1 2 ln + 1 (5) = ln 2 e nats. (6) = log 2 e bits. (7) (c) Sum of two normal distributions. The sum of two normal random variables is also normal, so applying the result derived the class for the normal distribution, since X 1 + X 2 N ( 1 + 2 , 2 1 + 2 2 ), h ( f ) = 1 2 log 2 e ( 2 1 + 2 2 ) bits. (8) 9.3 Mutual information for correlated normals. X Y N 2 , 2 2 2 2 (9) Using the expression for the entropy of a multivariate normal derived in class h ( X,Y ) = 1 2 log(2 e ) 2  K  = 1 2 log(2 e ) 2 4 (1 2 ) . (10) Since X and Y are individually normal with variance 2 , h ( X ) = h ( Y ) = 1 2 log 2 e 2 . (11) Hence I ( X ; Y ) = h ( X ) + h ( Y ) h ( X,Y ) = 1 2 log(1 2 ) . (12) 1 Solutions to problems from the text are supplied courtesy of Joy A. Thomas. 1 (a) = 1. In this case, X = Y , and knowing X implies perfect knowledge about Y . Hence the mutual information is infinite, which agrees with the formula. (b) = 0. In this case, X and Y are independent, and hence I ( X ; Y ) = 0, which agrees with the formula. (c) = 1. In this case, X = Y , and again the mutual information is infinite as in the case when = 1. 9.6 Scaling. Let Y = A X . Then the density of Y is g ( y ) = 1  A  f ( A 1 y ) . (13) Hence h ( A X ) = Z g ( y ) ln g ( y ) d y (14) = Z 1  A  f ( A 1 y ) ln f ( A 1 y ) log  A  d y (15) = Z 1  A  f ( x ) [ln f ( x ) log  A  ]  A  d x (16) = h ( X ) + log  A  . (17) 10.3 The two look Gaussian channel.The two look Gaussian channel....
View Full
Document
 Spring '10
 323

Click to edit the document details