Kybernetika 31 no. 4, 321-330, 1995

Statistical tests of optimality of source codes

Igor Vajda


For newly defined data compaction codes, as well as for the traditional data compression codes, we prove an asymptotic uniformity of probabilities of codewords, a kind of the asymptotic equipartition property. On the basis of this we propose an easily applicable Neyman-Pearson test of optimality of a code with a given asymptotic level of significance $0 < \alpha < 1$. The test is based on the sample entropy of code.


94B99, 94A99, 94A24, 62B10, 94A29