Rational Metareasoning and Compilation for Optimizing Decisions under Bounded Resources

From Tetherless World Wiki

Jump to: navigation, search

Citation: Eric Horvitz. (1989) Rational Metareasoning and Compilation for Optimizing Decisions under Bounded Resources. In KSL-89-81, 1989.

Publication techreport ( Edit )
type Technical Report
bibtype techreport
Bibtex basics
author Eric Horvitz
title Rational Metareasoning and Compilation for Optimizing Decisions under Bounded Resources
number KSL-89-81
institution Knowledge Systems, AI Laboratory
address Milan, Italy
year 1989
Bibtex more
Access Paper
abstract Several years ago, our research group initiated the Protos project to investigate the use of decision theory as a framework for reasoning about the design and operation of ideal agents under bounded resources. We have studied metareasoning, reasoning, and compilation within the framework of decision theory. Our model of rationality centers on the use of design-time and tractable run-time decision-theoretic analyses to control the detail and completeness of problem-level decision making. Unlike straightforward decision analyses, we apply the principles of decision theory to enriched models that include not only distinctions and outcomes in the world,but also distinctions and outcomes about cognition. In this paper, we shall review some earlier work on rational metareasoning and describe the benefits of integrating deliberative models of decision-theoretic reasoning and metareasoning with several classes of precomputed or compiled actions.

KSL Technical Report ID: KSL-89-81
Facts about Rational Metareasoning and Compilation for Optimizing Decisions under Bounded ResourcesRDF feed
Abstract Several years ago, our research group init Several years ago, our research group initiated the Protos project to investigate the use of decision theory as a framework for reasoning about the design and operation of ideal agents under bounded resources. We have studied metareasoning, reasoning, and compilation within the framework of decision theory. Our model of rationality centers on the use of design-time and tractable run-time decision-theoretic analyses to control the detail and completeness of problem-level decision making. Unlike straightforward decision analyses, we apply the principles of decision theory to enriched models that include not only distinctions and outcomes in the world,but also distinctions and outcomes about cognition. In this paper, we shall review some earlier work on rational metareasoning and describe the benefits of integrating deliberative models of decision-theoretic reasoning and metareasoning with several classes of precomputed or compiled actions. lasses of precomputed or compiled actions.
Address Milan, Italy  +
Author Eric Horvitz  +
Bibtype techreport  +
Has author Eric Horvitz  +
Has identifier KSL-89-81  +
Has publishing details 1989  +
Has title Rational Metareasoning and Compilation for Optimizing Decisions under Bounded Resources  +
Has where published KSL-89-81  +
Has year 1989  +
Institution Knowledge Systems, AI Laboratory  +
Ksl tr id KSL-89-81  +
Number KSL-89-81  +
Process note YES  +
Title Rational Metareasoning and Compilation for Optimizing Decisions under Bounded Resources  +
Year 1989  +
Personal tools