Convergent Dynamic Programming.
Abstract
Dynamic programming models are studied with finite total absolute return for each policy. It is shown that the supremum of the total expected return over the nearly conserving policies equals the supremum over all policies. A characterization is given of the existence of optimal policies. It is proved that the existence of an optimal policy implies the existence of a stationary optimal policy.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 30, 1974
- Accession Number
- ADA012902
Entities
People
- Arie Hordijk
Organizations
- Stanford University