By Topic

On the Discontinuity of the Shannon Information Measures

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
Siu-Wai Ho ; Dept. of Inf. Eng., Chinese Univ. of Hong Kong, Hong Kong, China ; Raymond W. Yeung

The Shannon information measures are well known to be continuous functions of the probability distribution for a given finite alphabet. In this paper, however, we show that these measures are discontinuous with respect to almost all commonly used "distance" measures when the alphabet is countably infinite. Such "distance" measures include the Kullback-Leibler divergence and the variational distance. Specifically, we show that all the Shannon information measures are in fact discontinuous at all probability distributions. The proofs are based on a probability distribution which can be realized by a discrete-time Markov chain with countably infinite number of states. Our findings reveal that the limiting probability distribution may not fully characterize the asymptotic behavior of a Markov chain. These results explain why certain existing information-theoretical tools are restricted to finite alphabets, and provide hints on how these tools can be extended to countably infinite alphabet.

Published in:

IEEE Transactions on Information Theory  (Volume:55 ,  Issue: 12 )