Some Bounds on Infinite Horizon Discounted Sequential Decision Processes.
Obtains new improved bounds on the optimal return function for many infinite horizon discounted sequential decision processes. The susceptible processes must satisfy what is called the "discounting" property. Processes satisfying the contraction and monotonicity properties qualify. Certain subprocesses are exploited, based on the simple notion of considering only relevant states. Some existing algorithms and some of their obvious extensions are listed. The possibility of identifying nonoptimal decisions is included. 22 pp. Ref.