Bayesian Representation of Stochastic Processes under Learning: de Finetti Revisited

A probability distribution governing the evolution of a stochastic process has infinitely many Bayesian representations of the form mu = integral operator [subscript theta] mu[subscript theta] delta lambda (theta). Among these, a natural representation is one whose components (mu[subscript theta]'s) are 'learnable' (one can approximate mu[subscript theta] by conditioning mu on observation of the process) and 'sufficient for prediction' (mu[subscript theta]'s predictions are not aided by conditioning on observation of the process). The authors show the existence and uniqueness of such a representation under a suitable asymptotic mixing condition on the process. This representation can be obtained by conditioning on the tail-field of the process, and any learnable representation that is sufficient for prediction is asymptotically like the tail-field representation. This result is related to the celebrated de Finetti theorem, but with exchangeability weakened to an asymptotic mixing condition, and with his conclusion of a decomposition into i.i.d. component distributions weakened to components that are learnable and sufficient for prediction.

[1]  D. Blackwell,et al.  Merging of Opinions with Increasing Information , 1962 .

[2]  A. Zellner.  An Introduction to Bayesian Inference in Econometrics , 1971 .

[3]  Meir Smorodinsky,et al.  Ergodic Theory Entropy , 1971 .

[4]  M. Rothschild.  A two-armed bandit theory of market pricing , 1974 .

[5]  David Blackwell,et al.  On Existence and Non-Existence of Proper, Regular, Conditional Distributions , 1975 .

[6]  P. Meyer,et al.  Probabilities and potential C , 1978 .

[7]  Mark Schilling,et al.  Goodness of Fit Testing in $\mathbb{R}^m$ Based on the Weighted Empirical Distribution of Certain Nearest Neighbor Statistics , 1983 .

[8]  S. Zamir,et al.  Formulation of Bayesian analysis for games with incomplete information , 1985 .

[9]  J. Szep,et al.  Games with incomplete information , 1985 .

[10]  Patrick Billingsley,et al.  Probability and Measure. , 1986 .

[11]  D. Freedman,et al.  On the consistency of Bayes estimates , 1986 .

[12]  Maxwell B. Stinchcombe.  Bayesian information topologies , 1990 .

[13]  James Bergin.  Player Type Distributions as State Variables and Information Revelation in Zero Sum Repeated Games with Discounting , 1992, Math. Oper. Res..

[14]  R. Rob,et al.  Learning, Mutation, and Long Run Equilibria in Games , 1993 .

[15]  T. Sargent.  Bounded rationality in macroeconomics , 1993 .

[16]  E. Kalai,et al.  Rational Learning Leads to Nash Equilibrium , 1993 .

[17]  Ehud Kalai,et al.  Weak and Strong Merging of Opinions , 1994 .

[18]  Doron Sonsino.  Learning to learn, pattern recognition, and Nash equilibrium , 1995 .

[19]  Robert J. Aumann,et al.  Repeated Games with Incomplete Information , 1995 .

[20]  Dov Samet,et al.  Common Priors and Markov Chains , 1996 .

[21]  R. Marimon.  Learning from learning in economics , 1996 .

[22]  Dov Samet.  Looking Backwards, Looking Inwards: Priors and Introspection , 1996 .

[23]  Aggregation and the Law of Large Numbers in Economies with a Continuum of Agents , 1996 .

[24]  M. Jackson,et al.  Social Learning in Recurring Games , 1997 .

[25]  Ehud Lehrer,et al.  Repeated Large Games with Incomplete Information , 1997 .

[26]  Yaw Nyarko.  Bayesian learning and convergence to Nash equilibria without common priors , 1998 .

[27]  Nabil I. Al-Najjar.  FACTOR ANALYSIS AND ARBITRAGE PRICING IN LARGE ASSET ECONOMIES , 1998 .

[28]  D. Fudenberg,et al.  Conditional Universal Consistency , 1999 .