Algorithms for distribution matching are considered that approximate discrete memoryless sources by transforming a sequence of coin-flipping bits to a sequence of independent and identically distributed symbols with a desired distribution. Two fixed-to-fixed length code book constructions are studied. The code classes are constant composition codes, i.e., codes that consist only of permutations of one word, and codes that approximate discrete memoryless sources best in terms of informational divergence.
«
Algorithms for distribution matching are considered that approximate discrete memoryless sources by transforming a sequence of coin-flipping bits to a sequence of independent and identically distributed symbols with a desired distribution. Two fixed-to-fixed length code book constructions are studied. The code classes are constant composition codes, i.e., codes that consist only of permutations of one word, and codes that approximate discrete memoryless sources best in terms of informational div...
»