• DocumentCode
    2620092
  • Title

    Prefixes and the entropy rate for long-range sources

  • Author

    Kontoyiannis, Ioannis ; Suhov, Yurii M.

  • Author_Institution
    Inf. Syst. Lab., Stanford Univ., CA, USA
  • fYear
    1994
  • fDate
    27 Jun-1 Jul 1994
  • Firstpage
    194
  • Abstract
    The asymptotic a.s.-relation H=limn→∞[(nlogn)/(Σi=1n Lin(X))] is derived for any finite-valued stationary ergodic process X=(Xn, n∈Z) that satisfies a Doeblin-type condition: there exists r⩾1 such that essxinf P(Xn+1|x→∞,n)⩾α>0. Here, H is the entropy rate of the process X, and Lin(X) is the length of a shortest prefix in X which is initiated at time i and is not repeated among the prefixes initiated at times j, 1⩽i≠J⩽n. The validity of this limiting result was established by Shields in 1989 for i.i.d. processes and also for irreducible aperiodic Markov chains. Under our new condition, we prove that this holds for a wider class of processes, that may have infinite memory
  • Keywords
    entropy codes; source coding; Doeblin-type condition; IID processes; entropy rate; finite-valued stationary ergodic process; infinite memory; irreducible aperiodic Markov chains; long-range sources; prefixes; Entropy; Estimation theory; Information systems; Laboratories; Stochastic processes; Writing;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Information Theory, 1994. Proceedings., 1994 IEEE International Symposium on
  • Conference_Location
    Trondheim
  • Print_ISBN
    0-7803-2015-8
  • Type

    conf

  • DOI
    10.1109/ISIT.1994.394774
  • Filename
    394774