Shows that a discrete infinite distribution with finite entropy cannot be estimated consistently in information divergence. As a corollary the authors show that there is no universal source code for an infinite source alphabet over the class of all discrete memoryless sources with finite entropy.
展开▼
