An efficient universal prediction algorithm for unknown sources with limited training data
J Ziv - IEEE Transactions on Information Theory, 2002 - ieeexplore.ieee.org
Inspired by CE Shannon's celebrated paper:" Prediction and entropy of printed
English"(1951), we consider the optimal prediction error for unknown finite-alphabet ergodic
Markov sources, for prediction algorithms that make inference about the most probable
incoming letter, where the distribution of the unknown source is apparent only via a short
training sequence of N+ 1 letters. We allow N to be a polynomial function of K, the order of
the Markov source, rather than the classical case where N is allowed to be exponential in K …
English"(1951), we consider the optimal prediction error for unknown finite-alphabet ergodic
Markov sources, for prediction algorithms that make inference about the most probable
incoming letter, where the distribution of the unknown source is apparent only via a short
training sequence of N+ 1 letters. We allow N to be a polynomial function of K, the order of
the Markov source, rather than the classical case where N is allowed to be exponential in K …
Showing the best result for this search. See all results
