6 of 7 people found the following review helpful
on August 5, 2009
A colleague recently opined "You don't hear much about Information Theory any more". I disagree with the sentiment. I think the problem is that I.T. has gone from the philosophical and the curious to something with applications, convertible to problem sets!
Yeung's contribution is thoroughly modern, beginning with Shannon and Markov through divergence and other measures. Traditional memoryless channels aren't neglected, nor is the famous Blahut and Arimoto rate distortion problem frame.
Connections are made throughout, whether to the EM algorithm (Dempster, Laird, Rubin) or implicit applications of notions like divergence to genomics and ecological modeling.
Of personal interest is the strong section on network coding, arguing why additional coding is needed at intermediate nodes to maintain bandwidth optimality, despite coding at sources and for the channel.