An adaptive universal variable-to-fix lossless coding scheme

We suggest an adaptive, universal variable-to-fix (VF) lossless source coding technique which generates a dictionary that is sequentially adapted to the empirical distribution of the data observed so far. Since the dictionary varies, we also provide an algorithm to implement the VF code without an explicit specification of the dictionary, and so there is no need for dictionary storage. For i.i.d sources our algorithm converges to the entropy at a rate O(l/logM)+O(logn/n), where M is the dictionary size and a is the length of the sequence to be compressed. The adaptive dictionary construction algorithm has some resemblance to arithmetic coding