Learning the ergodic decomposition

This addresses the theoretical problem of Bayesian learning in ergodic theory for statisticians and machine learning researchers, but it appears incremental as it builds on established concepts like ergodic decomposition.

The paper tackles the problem of a Bayesian agent learning the structure of a stationary process from past observations, proving that the agent's predictions about the near future converge to those based on the long-run empirical frequencies of the process.