Abstract
The problem of calculating the rate of mutual information between two coarse-grained variables that together specify a continuous time Markov process is addressed. As a main obstacle, the coarse-grained variables are in general non-Markovian, therefore, an expression for their Shannon entropy rates in terms of the stationary probability distribution is not known. A numerical method to estimate the Shannon entropy rate of continuous time hidden-Markov processes from a single time series is developed. With this method the rate of mutual information can be determined numerically. Moreover, an analytical upper bound on the rate of mutual information is calculated for a class of Markov processes for which the transition rates have a bipartite character. Our general results are illustrated with explicit calculations for four-state networks.
Similar content being viewed by others
References
Shannon, C.E.: Bell Syst. Tech. J. 27, 379–423 (1948)
Cover, T.M., Thomas, J.A.: Elements of Information Theory, 2nd edn. Wiley, Hoboken (2006)
Li, W.: J. Stat. Phys. 60, 823 (1990)
Barato, A.C., Hartich, D., Seifert, U.: Phys. Rev. E 87, 042104 (2013)
Holliday, T., Goldsmith, A., Glynn, P.: IEEE Trans. Inf. Theory 52, 3509 (2006)
Jacquet, P., Seroussi, G., Szpankowski, W.: Theor. Comput. Sci. 395, 203 (2008)
Roldan, E., Parrondo, J.M.R.: Phys. Rev. E 85, 031129 (2012)
Seifert, U.: Rep. Prog. Phys. 75, 126001 (2012)
Touchette, H., Lloyd, S.: Phys. Rev. Lett. 84, 1156 (2000)
Cao, F.J., Feito, M.: Phys. Rev. E 79, 041118 (2009)
Sagawa, T., Ueda, M.: Phys. Rev. Lett. 104, 090602 (2010)
Sagawa, T., Ueda, M.: Phys. Rev. Lett. 109, 180602 (2012)
Sagawa, T., Ueda, M.: Phys. Rev. E 85, 021104 (2012)
Horowitz, J.M., Vaikuntanathan, S.: Phys. Rev. E 82, 061120 (2010)
Horowitz, J.M., Parrondo, J.M.R.: Europhys. Lett. 95(1), 10005 (2011)
Granger, L., Kantz, H.: Phys. Rev. E 84, 061110 (2011)
Abreu, D., Seifert, U.: Phys. Rev. Lett. 108, 030601 (2012)
Abreu, D., Seifert, U.: Europhys. Lett. 94, 10001 (2011)
Bauer, M., Abreu, D., Seifert, U.: J. Phys. A, Math. Theor. 45, 162001 (2012)
Ito, S., Sano, M.: Phys. Rev. E 84, 021123 (2011)
Crisanti, A., Puglisi, A., Villamaina, D.: Phys. Rev. E 85, 061127 (2012)
Mandal, D., Jarzynski, C.: Proc. Natl. Acad. Sci. USA 109, 11641 (2012)
Esposito, M., Schaller, G.: Europhys. Lett. 99, 30003 (2012)
Strasberg, P., Schaller, G., Brandes, T., Esposito, M.: Phys. Rev. Lett. 110, 040601 (2013)
Barato, A.C., Seifert, U.: Europhys. Lett. 101, 60001 (2013)
Lan, G., Sartori, P., Neumann, S., Sourjik, V., Tu, Y.: Nat. Phys. 8, 422 (2012)
Mehta, P., Schwab, D.J.: Proc. Natl. Acad. Sci. USA 109, 17978 (2012)
Ephraim, Y., Merhav, N.: IEEE Trans. Inf. Theory 48, 1518 (2002)
Gaspard, P.: J. Stat. Phys. 117, 599 (2004)
Lecomte, V., Appert-Rolland, C., Wijland, F.: J. Stat. Phys. 127, 51 (2007)
Dumitrescu, M.B.: Čas. Pěst. Mat. 113, 429 (1988)
Schnakenberg, J.: Rev. Mod. Phys. 48, 571 (1976)
Kawai, R., Parrondo, J.M.R., van den Broeck, C.: Phys. Rev. Lett. 98, 080602 (2007)
Rahav, S., Jarzynski, C.: J. Stat. Mech., Theor. Exp. P09012 (2007)
Esposito, M.: Phys. Rev. E 85, 041125 (2012)
Crisanti, A., Paladin, G., Vulpiani, A.: Products of Random Matrices in Statistical Physics. Springer Series in Solid State Sciences (1993)
Acknowledgements
Support by the ESF through the network EPSD is gratefully acknowledged.
Author information
Authors and Affiliations
Corresponding author
Appendix: Detailed Derivation of the Analytical Upper Bound
Appendix: Detailed Derivation of the Analytical Upper Bound
The first upper bound H(Y 2|Y 1) can be easily calculated by using the conditional probability
where Y 2≠Y 1. We here performed the substitutions X 1→α, Y 1→i, and Y 2→j. Using this formula in (22) we obtain
Moreover, H(Y N+1|Y N ,…,Y 1) up to order τ is given by the above formula for any finite N. In order to demonstrate this we first rewrite (22) as
where P(Y N ,Y N ,…,Y 1) denotes the probability of having a sequence for which Y N+1=Y N . For Y N+1≠Y N , the expression of the conditional probability P(Y N+1|Y N ,…,Y 1) has at least one transition probability term of order τ. Therefore, as P(Y N+1|Y N ,…,Y 1) is at least a term of order τ, it is convenient to further rewrite the above expression as
where in the first line we summed over the variables Y 1,…,Y N−1. The three following relations are important for the subsequent derivation. First, for Y N+1≠Y N ,
Moreover,
where η≥1 is an integer and A is a constant independent of τ. Finally, the conditional probability distribution fulfills
where ν≥1 is an integer and B is a constant independent of τ. With these three relations, the term in the second line in Eq. (71) becomes
where we used τ ν+η−1lnτ ν−1∈O(τ). For the term in the third line in Eq. (71) we need the relations,
and
which lead to
Inserting (75) and (78) in (71) we obtain
Therefore, since the Y process is stationary, from (69), we obtain for any finite N
Applying the same method to the X process we get,
Rights and permissions
About this article
Cite this article
Barato, A.C., Hartich, D. & Seifert, U. Rate of Mutual Information Between Coarse-Grained Non-Markovian Variables. J Stat Phys 153, 460–478 (2013). https://doi.org/10.1007/s10955-013-0834-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10955-013-0834-5