On Incomplete Learning and Certainty-Equivalence Control

Operations Research, Vol. 66, No. 4, July-August 2018, pp. 1136-1167

Columbia Business School Research Paper No. 17-63

57 Pages Posted: 12 Jun 2017 Last revised: 1 Oct 2018

See all articles by N. Bora Keskin

N. Bora Keskin

Duke University - Fuqua School of Business

Assaf Zeevi

Columbia University - Columbia Business School, Decision Risk and Operations

Date Written: November 17, 2017

Abstract

We consider a dynamic learning problem where a decision maker sequentially selects a control and observes a response variable that depends on chosen control and an unknown sensitivity parameter. After every observation, the decision maker updates her/his estimate of the unknown parameter and uses a certainty-equivalence decision rule to determine subsequent controls based on this estimate. We show that under this certainty-equivalence learning policy the parameter estimates converge with positive probability to an uninformative fixed point that can differ from the true value of the unknown parameter; a phenomenon that will be referred to as incomplete learning. In stark contrast, it will be shown that this certainty-equivalence policy may avoid incomplete learning if the parameter value of interest “drifts away” from the uninformative fixed point at a critical rate. Finally, we prove that one can adaptively limit the learning memory to improve the accuracy of the certainty-equivalence policy in both static (estimation), as well as slowly varying (tracking) environments, without relying on forced exploration.

Keywords: Dynamic Control, Sequential Estimation, Certainty Equivalence, Incomplete Learning

Suggested Citation

Keskin, N. Bora and Zeevi, Assaf, On Incomplete Learning and Certainty-Equivalence Control (November 17, 2017). Operations Research, Vol. 66, No. 4, July-August 2018, pp. 1136-1167, Columbia Business School Research Paper No. 17-63, Available at SSRN: https://ssrn.com/abstract=2984088 or http://dx.doi.org/10.2139/ssrn.2984088

N. Bora Keskin (Contact Author)

Duke University - Fuqua School of Business ( email )

100 Fuqua Drive
Durham, NC 27708-0120
United States

HOME PAGE: http://faculty.fuqua.duke.edu/~nk145/

Assaf Zeevi

Columbia University - Columbia Business School, Decision Risk and Operations ( email )

New York, NY
United States
212-854-9678 (Phone)
212-316-9180 (Fax)

HOME PAGE: http://www.gsb.columbia.edu/faculty/azeevi/

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
168
Abstract Views
1,227
Rank
378,371
PlumX Metrics