Asymptotic Properties of Learning Models

  • Marius Iosifescu
Conference paper
Part of the Lecture Notes in Statistics book series (LNS, volume 20)


It is already well known that all stochastic models for learning studied up to now enter the following general scheme (cf.Norman (1972)). The behaviour of the subject on trial n = 0,1,... is determined by his state Sn (an indicator of the subject’s response tendencies) at the beginning of the trial. Sn is a random variable taking on values in a measurable state space (S,S). On trial n an event En +1 occurs that results in a change of state. En +1 is a randon variable taking on values in a measurable event space (E,E) and specifies those occurrences on trial n that affect subsequent behaviour. Typically, En +1. includes a specification of the subject’s response and its observable outcome or payoff. To represent the fact that the occurrence of an event effects a change of state we consider a measurable mapping v of S × E into S and postulate that Sn +1 = v(Sn, En +1), n ≥ 0. Finally, we assume that the conditional probability distribution of En +1 given En, Sn,... depends only on the state Sn and denote it Q(Sn,.).


Iterate Logarithm Conditional Probability Distribution Random System Stationary Increment Functional Central Limit Theorem 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. Billingsley, P.(1968) Convergence of Probability Measures. New York: Wiley.MATHGoogle Scholar
  2. Botkin, J.W., Elmandjra, M. and Malitza, M. (1980) No Limits to Learning. Bridging the Human Gap: A Report fco the Club of Rome. Oxford: Pergamon Press.Google Scholar
  3. Grigorescu, S. and Iosifescu, M. (1983) Dependence with Complete Connections and Its Applicatcions. Bucharest: Scientific and Encyclopaedic Publishing House.(In Romanian)Google Scholar
  4. Heyde, C.C. and Scott, D.J.(1973) Invariance principles for the law of the iterated logarithm for martingales and processes with stationary increments. Ann.Probab. 1, 428–436.MathSciNetMATHCrossRefGoogle Scholar
  5. Iosifescu, M.(1963) Random systems with complete connections with an arbitrary set of states. Rev.Math. Pures Appl. 8, 611–645;MathSciNetGoogle Scholar
  6. Iosifescu, M.(1963) Random systems with complete connections with an arbitrary set of states. Rev.Math. Pures Appl. 9(1964), 91–92.(In Russian)MathSciNetGoogle Scholar
  7. Iosifescu, M.(l972) On Strassen’s version of the loglog law for some classes of dependent random variables. Z.Wahrsch.Verw.Gebiete 24, 155–158.MathSciNetCrossRefGoogle Scholar
  8. Iosifescu, M. and Theodorescu, R.(1969) Random Processes and Learning. Berlin: Springer.MATHGoogle Scholar
  9. Marliss, G.S. and McGregor, J.R.(1971) The construction of limiting distributions of response probabilities. J.Appl.Probab. 8, 757–766.MathSciNetMATHCrossRefGoogle Scholar
  10. Norman, M.F.(1972) Markov Processes and Learning Models. New York: Academic Press.MATHGoogle Scholar
  11. Popescu, Gh.(1976) A functional central limit theorem for a class of Markov chains. Rev.Roumaine Math.Pures Appl. 21, 737–750.MathSciNetMATHGoogle Scholar
  12. Popescu, Gh.(1977) Functional limit theorems for random systems with complete connections. Proc.Fifth Conf.Probability Theory (Braşov, 1974), pp.261–275. Bucharest: Publishing House of the Romanian Academy.Google Scholar
  13. Scott, D.J.(1971) Central limit theorems for martingales and for processes with stationary increments, using a Skorokhod representation approach. Appl.Probab. 5, 119–137.CrossRefGoogle Scholar
  14. Zamfirescu, T. (198l) Most monotone functions are singular. Amer.Math. Monthly 88, 47–49.MathSciNetCrossRefGoogle Scholar

Copyright information

© Springer-Verlag New York Inc. 1983

Authors and Affiliations

  • Marius Iosifescu
    • 1
  1. 1.Centre of Mathematical StatisticsBucharestRomania

Personalised recommendations