The information content of binary sequences is defined by minimal program complexity measures and is related to computable martingales. The equivalence of the complexity approach and the martingale approach after restriction to effective random tests is used to establish generalized source coding theorems and converses. Finite state complexity and decomposable martingales are related to classical block codes and the relative frequency behavior of sequences.
Published in:
Information Theory, IEEE Transactions on
(Volume:25
,
Issue:
5
)
Date of Publication: Sep 1979