lecture9 notes

# Y x e x x x x px x note also that xx px x

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: n � � P (y |x) log(PX |Y (x|y )) e y∈Y Y |X �� � PX (x) = � � ) log(P �X |Y (x�|y)) PY |X (y |x y ∈Y x� ∈X e � � � PY |X (y |x) PY |X (y |x) log � y ∈Y PX (x� )PY |X (y |x� ) x� ∈X PX (x)e ⎛ = � �⎞ � PY |X (y |x) PY |X (y |x� ) log � ⎜ ⎟ y ∈Y � PX (x� )PY |X (y |x� ) � ) ⎜e ⎟ x� ∈X x� ∈X PX (x ⎝ ⎠ Note also that � x∈X PX (x) = 1. This may b e very hard to solve. Arimoto-Blahut Proof: The ﬁrst two statements follow immedi­ ately from our lemma For any value of x where PX |Y (x|y ) = 0, PX (x) should b e set to 0 to obtain the max­ imum. To ﬁnd the maximum over the PMF PX , let us ﬁrst ignore the constraint of p ositiv­ ity and use a Lagrange multiplier for the � x PX (x) = 1 Then ∂ � � { x∈X y∈Y PX (x)PY |X (y |x) log ∂ PX (x) � λ ( x∈X PX (x) − 1)} = 0 � � �X |Y (x|y ) P + PX (x) Arimoto-Blahut Equivalently � � � � − log(PX (x))−1+ y∈Y PY |X (y |x) lo...
View Full Document

## This document was uploaded on 03/19/2014 for the course EECS 6.441 at MIT.

Ask a homework question - tutors are online