## Reasoning about Beliefs and Actions under Computational Resource Constraints (1987)

### Cached

### Download Links

- [research.microsoft.com]
- [ftp.research.microsoft.com]
- DBLP

### Other Repositories/Bibliography

Venue: | In Proceedings of the 1987 Workshop on Uncertainty in Artificial Intelligence |

Citations: | 191 - 20 self |

### BibTeX

@INPROCEEDINGS{Horvitz87reasoningabout,

author = {Eric J. Horvitz},

title = {Reasoning about Beliefs and Actions under Computational Resource Constraints},

booktitle = {In Proceedings of the 1987 Workshop on Uncertainty in Artificial Intelligence},

year = {1987},

pages = {429--444}

}

### Years of Citing Articles

### OpenURL

### Abstract

ion Modulation In many cases, it may be more useful to do normative inference on a model that is deemed to be complete at a particular level of abstraction than it is to do an approximate or heuristic analysis of a model that is too large to be analyzed under specific resource constraints. It may prove useful in many cases to store several belief-network representations, each containing propositions at different levels of abstraction. In many domains, models at higher levels of abstraction are more tractable. As the time available for computation decreases, network modules of increasing abstraction can be employed. ffl Local Reformulation Local reformulation is the modification of specific troublesome topologies in a belief network. Approximation methods and heuristics designed to modify the microstructure of belief networks will undoubtedly be useful in the tractable solution of large uncertain-reasoning problems. Such strategies might be best applied at knowledge-encoding time. An...

### Citations

11403 |
Computers and Intractability: A Guide to the Theory of NP-Completeness
- Garey, Johnson
- 1979
(Show Context)
Citation Context ...l answers permeates computer science. Complexity theorists have focused almost exclusively on proving results about the time and space resources that must be expended to run algorithms to termination =-=[8, 1, 25]-=-. In the real world, strict limitations and variations on the time available for problem solving suggest that the focus on time complexity for algorithmic termination is limited; analyses centering on... |

2424 |
A Mathematical Theory of Evidence
- Shafer
- 1976
(Show Context)
Citation Context ...esence of a hypothesis can range continuously between complete truth and complete falsity. Such beliefentailment schemes include probability theory [18, 27], fuzzy logic [38], Dempster--Shafer theory =-=[30]-=-, and MYCIN certainty factors [2]. Finally, decision making is the process of selecting the best action to take. A decision or action is an irrevocable allocation of valuable resources. The classical ... |

1337 |
Combinatorial optimization: Algorithms and complexity
- Papadimitriou, Steiglitz
- 1982
(Show Context)
Citation Context ...l answers permeates computer science. Complexity theorists have focused almost exclusively on proving results about the time and space resources that must be expended to run algorithms to termination =-=[8, 1, 25]-=-. In the real world, strict limitations and variations on the time available for problem solving suggest that the focus on time complexity for algorithmic termination is limited; analyses centering on... |

1152 |
The Theory of Games and Economic Behavior
- Neumann, Morgenstern
- 1944
(Show Context)
Citation Context ...istent with a parsimonious set of axioms. The logical equivalence of these axioms with a small set of intuitive properties desired in a measure of belief has been demonstrated [7, 18]. Utility theory =-=[36]-=- dictates the consistent assignment and updating of the value of alternative actions given the values of alternative outcomes and the degrees of belief in the outcomes. Measures of value consistent wi... |

872 |
A Behavioral Model of Rational Choice
- Simon
- 1955
(Show Context)
Citation Context ...ar problem-solving challenges. Notions of bounded rationality that have been exercised in earlier discussion of intelligent systems faced with complex problems shun a formal perspective as too costly =-=[31, 23]-=-. Most research on reasoning and acting under resource constraints has focused on the discovery of relatively simple satisficing approaches to problem solving. These approaches may stray far from the ... |

767 |
Data Structures and Algorithms
- Aho, Hopcroft, et al.
- 1983
(Show Context)
Citation Context ...l answers permeates computer science. Complexity theorists have focused almost exclusively on proving results about the time and space resources that must be expended to run algorithms to termination =-=[8, 1, 25]-=-. In the real world, strict limitations and variations on the time available for problem solving suggest that the focus on time complexity for algorithmic termination is limited; analyses centering on... |

398 |
Evaluating influence diagrams
- Shachter
- 1986
(Show Context)
Citation Context ... from the National Library of Medicine. Computer facilities were provided by the SUMEX-AIM Resource under Grant RR-00785 from the National Institutes of Health. relatively complex real-world problems =-=[29, 13]-=-. In this paper, we move beyond discussions of the degree to which the theories of probability and utility are able to solve real-world problems. We focus on situations where it is clear that insuffic... |

395 | Fusion, propagation and structuring in belief networks
- Pearl
- 1986
(Show Context)
Citation Context ...ost schemes, the degree of truth or belief in the presence of a hypothesis can range continuously between complete truth and complete falsity. Such beliefentailment schemes include probability theory =-=[18, 27]-=-, fuzzy logic [38], Dempster--Shafer theory [30], and MYCIN certainty factors [2]. Finally, decision making is the process of selecting the best action to take. A decision or action is an irrevocable ... |

319 |
Rule-based expert systems : the MYCIN experiments of the Stanford Heuristic Programming Project
- Buchanan, Shortliffe
- 1984
(Show Context)
Citation Context ...ontinuously between complete truth and complete falsity. Such beliefentailment schemes include probability theory [18, 27], fuzzy logic [38], Dempster--Shafer theory [30], and MYCIN certainty factors =-=[2]-=-. Finally, decision making is the process of selecting the best action to take. A decision or action is an irrevocable allocation of valuable resources. The classical decision-theoretic basis defines ... |

177 |
Probability, frequency and reasonable expectation
- Cox
- 1946
(Show Context)
Citation Context ...positions should be consistent with a parsimonious set of axioms. The logical equivalence of these axioms with a small set of intuitive properties desired in a measure of belief has been demonstrated =-=[7, 18]-=-. Utility theory [36] dictates the consistent assignment and updating of the value of alternative actions given the values of alternative outcomes and the degrees of belief in the outcomes. Measures o... |

150 |
Bounded Rationality, Ambiguity and the Engineering of Choice
- March
- 1978
(Show Context)
Citation Context ...ar problem-solving challenges. Notions of bounded rationality that have been exercised in earlier discussion of intelligent systems faced with complex problems shun a formal perspective as too costly =-=[31, 23]-=-. Most research on reasoning and acting under resource constraints has focused on the discovery of relatively simple satisficing approaches to problem solving. These approaches may stray far from the ... |

99 |
Evidential Reasoning Using Stochastic Simulation of Causal Models
- Pearl
- 1987
(Show Context)
Citation Context ...chniques are approximation strategies that report a probability distribution or partial characterization of a distribution over probabilities of interest through a process of weighted random sampling =-=[13, 26]-=-. In many cases, the distribution over the probabilities is approximated by the binomial distribution. The variance with which the distribution converges on a probability with additional computation d... |

98 |
Probabilistic Inference Using Belief Networks is NP-Hard
- Cooper
- 1987
(Show Context)
Citation Context ...esisted tractable algorithmic solution. A troubling topology is the multiply connected network [27]. Such inference problems belong to a class of difficult problems that have been shown to be NP-hard =-=[6]-=-. Problems in complex areas such as medicine often require representation with multiply connected networks. Thus, in the worst case, rational beliefs and actions demand computation that is exponential... |

95 |
The role of fuzzy logic in the management of uncertainty in expert systems
- Zadeh
- 1983
(Show Context)
Citation Context ...e of truth or belief in the presence of a hypothesis can range continuously between complete truth and complete falsity. Such beliefentailment schemes include probability theory [18, 27], fuzzy logic =-=[38]-=-, Dempster--Shafer theory [30], and MYCIN certainty factors [2]. Finally, decision making is the process of selecting the best action to take. A decision or action is an irrevocable allocation of valu... |

88 |
Probabilistic interpretations for MYCINâ€™s certainty factors
- Heckerman
- 1986
(Show Context)
Citation Context ...rational beliefs and actions. Numerous investigators interested in the automation of uncertain reasoning have converged on the theoretical adequacy of the decision-theoretic basis for rational action =-=[11, 3, 17]-=-. Recent discussions about computational approaches to reasoning with uncertainty have centered on the degree to which probability and utility theory can handle inference problems of realistic complex... |

83 | defence of probability
- In
- 1985
(Show Context)
Citation Context ...rational beliefs and actions. Numerous investigators interested in the automation of uncertain reasoning have converged on the theoretical adequacy of the decision-theoretic basis for rational action =-=[11, 3, 17]-=-. Recent discussions about computational approaches to reasoning with uncertainty have centered on the degree to which probability and utility theory can handle inference problems of realistic complex... |

65 |
Rational, Descriptions, Decisions and Designs
- Tribus
- 1969
(Show Context)
Citation Context ...rrounding the formulation of problems [18]. Other researchers have shown that probability theory and utility theory are logically equivalent to the satisfaction of a small set of intuitive properties =-=[35, 18]-=-. Still others have responded to complaints of intractability by demonstrating techniques that can solve 1 This article appeared originally in the Proceedings of the Third Workshop on Uncertainty in A... |

41 |
Fuzzy Decision Analysis
- Watson, Weiss, et al.
- 1979
(Show Context)
Citation Context ... of doing a decision analysis or continuing to refine a decision. Matheson [24] explored the value of spending additional effort to analyze a bidding problem proposed by Howard [20]. Watson and Brown =-=[37]-=- and Lindley [22] describe issues surrounding the application of a preliminary decision analysis to assist in the decision as to whether an individual should embark on a more costly decision analysis ... |

35 |
SG: Categorical and probabilistic reasoning in medical diagnosis
- Szolovits, Pauker
(Show Context)
Citation Context ...obal independence have been made in many reasoning systems that have been deemed to perform adequately (e.g., the MYCIN certainty-factor model [11, 17] and the early probabilistic diagnostic programs =-=[10, 33]-=-). The actual costs and benefits of defaulting to conditional independence among evidence in many real-world problems have not been determined. A promising area of research is the prioritization of cl... |

28 |
Propagation of uncertainty by probabilistic logic sampling in bayesâ€™ networks
- HENRION
- 1988
(Show Context)
Citation Context ... from the National Library of Medicine. Computer facilities were provided by the SUMEX-AIM Resource under Grant RR-00785 from the National Institutes of Health. relatively complex real-world problems =-=[29, 13]-=-. In this paper, we move beyond discussions of the degree to which the theories of probability and utility are able to solve real-world problems. We focus on situations where it is clear that insuffic... |

23 |
A Computer-based Medical Diagnostic Aid that Integrates Causal and Probabilistic Knowledge
- NESTOR
- 1984
(Show Context)
Citation Context ...tain-reasoning problems represented with graphs. The most popular representation uses directed graphs to represent explicitly conditional dependencies and independencies among beliefs in propositions =-=[28, 5, 27]-=-. Many researchers have ascribed a common semantics to the directed graphs. The representation is often called a belief network. In a belief network, an arc between a node representing proposition A a... |

23 |
Evaluating In Diagrams
- Shachter
- 1986
(Show Context)
Citation Context ... from the National Library of Medicine. Computer facilities were provided by the SUMEX-AIM Resource under Grant RR-00785 from the National Institutes of Health. relatively complex real-world problems =-=[29, 13]-=-. In this paper, we move beyond discussions of the degree to which the theories of probability and utility are able to solve real-world problems. We focus on situations where it is clear that insucie... |

20 |
The limits of inference
- Smith
- 1955
(Show Context)
Citation Context ...gence research on the control of reasoning by Smith, and by Treitel and Genesereth, has explored the usefulness of applying utility theory in the selection of alternative logical reasoning strategies =-=[32, 34]-=-. Our work differs from the logic theorem-proving work in its pursuit of theoretical tools for the control of computation under scarce resources, of decision-theoretic inference under resource constra... |

19 |
The Economic Value of Analysis and Computation
- Matheson
- 1968
(Show Context)
Citation Context ...ces. Other related work in decision science centered on the value of analysis. These studies have explored the likely benefit of doing a decision analysis or continuing to refine a decision. Matheson =-=[24]-=- explored the value of spending additional effort to analyze a bidding problem proposed by Howard [20]. Watson and Brown [37] and Lindley [22] describe issues surrounding the application of a prelimin... |

18 |
Value of information lotteries
- Howard
- 1967
(Show Context)
Citation Context ...ored the likely benefit of doing a decision analysis or continuing to refine a decision. Matheson [24] explored the value of spending additional effort to analyze a bidding problem proposed by Howard =-=[20]-=-. Watson and Brown [37] and Lindley [22] describe issues surrounding the application of a preliminary decision analysis to assist in the decision as to whether an individual should embark on a more co... |

17 | A framework for comparing alternative formalisms for plausible reasoning
- Horvitz, Heckerman, et al.
- 1986
(Show Context)
Citation Context ...expressiveness of probability theory by pointing out that the normative basis addresses consistent inference with measures of belief and preference, not issues surrounding the formulation of problems =-=[18]-=-. Other researchers have shown that probability theory and utility theory are logically equivalent to the satisfaction of a small set of intuitive properties [35, 18]. Still others have responded to c... |

15 |
The inconsistent use of measures of certainty in artificial intelligence research
- Horvitz, Heckerman
- 1990
(Show Context)
Citation Context ...rational beliefs and actions. Numerous investigators interested in the automation of uncertain reasoning have converged on the theoretical adequacy of the decision-theoretic basis for rational action =-=[11, 3, 17]-=-. Recent discussions about computational approaches to reasoning with uncertainty have centered on the degree to which probability and utility theory can handle inference problems of realistic complex... |

14 |
Bayesian belief network inference using simulation
- Chin, Cooper
- 1989
(Show Context)
Citation Context ...network, and on the nature of the probabilistic dependencies within the network. Recent work has shown current simulation algorithms to have intolerably slow convergence rates in many realistic cases =-=[4]-=- . Stochastic simulation is nevertheless a promising class of inference for the derivation of useful bounded-resource computation strategies. ffl Completeness Modulation Completeness-modulation strate... |

11 |
Computer-assisted clinical decision making
- Gorry
- 1973
(Show Context)
Citation Context ...obal independence have been made in many reasoning systems that have been deemed to perform adequately (e.g., the MYCIN certainty-factor model [11, 17] and the early probabilistic diagnostic programs =-=[10, 33]-=-). The actual costs and benefits of defaulting to conditional independence among evidence in many real-world problems have not been determined. A promising area of research is the prioritization of cl... |

9 |
On the expressiveness of rule-based systems for reasoning under uncertainty
- Heckerman, Horvitz
- 1987
(Show Context)
Citation Context ... not directly dependent on the values of A. Less expressive representations commonly employed in artificialintelligence research have not allowed specific independencies to be represented efficiently =-=[12]-=-. 1 Belief networks are special cases of more general graphical representations that can represent availablesactions and the utility of alternative outcomes in addition to beliefs [21, 29]. These grap... |

8 |
Choosing directions for rules
- Treitel, Genesereth
- 1986
(Show Context)
Citation Context ...gence research on the control of reasoning by Smith, and by Treitel and Genesereth, has explored the usefulness of applying utility theory in the selection of alternative logical reasoning strategies =-=[32, 34]-=-. Our work differs from the logic theorem-proving work in its pursuit of theoretical tools for the control of computation under scarce resources, of decision-theoretic inference under resource constra... |

6 |
A multiattribute utility approach to inference understandability and explanation
- Horvitz
- 1987
(Show Context)
Citation Context ...e delay and deadlines. The node Clarity indicates that the transparency and explainability of inference may be another important dimension of inference-related cost in the context of an expert system =-=[19, 16]-=-. 5.2 Multiple Attributes of Inference and Outcome As indicated in Figure 2, the costs of reasoning may have several components. Multiple dimensions of utility can be ascertained through consideration... |

5 |
A method for computing probabilities in complex situations
- Rousseau
- 1968
(Show Context)
Citation Context ...tain-reasoning problems represented with graphs. The most popular representation uses directed graphs to represent explicitly conditional dependencies and independencies among beliefs in propositions =-=[28, 5, 27]-=-. Many researchers have ascribed a common semantics to the directed graphs. The representation is often called a belief network. In a belief network, an arc between a node representing proposition A a... |

4 |
The use of a heuristic problemsolving hierarchy to facilitate the explanation of hypothesis-directed reasoning
- Horvitz, Heckerman, et al.
- 1986
(Show Context)
Citation Context ...e delay and deadlines. The node Clarity indicates that the transparency and explainability of inference may be another important dimension of inference-related cost in the context of an expert system =-=[19, 16]-=-. 5.2 Multiple Attributes of Inference and Outcome As indicated in Figure 2, the costs of reasoning may have several components. Multiple dimensions of utility can be ascertained through consideration... |

3 |
Toward a science of expert systems
- Horvitz
- 1986
(Show Context)
Citation Context ...ing such resource insufficiencies. We have been exploring resource constraints and the feasibility of normative approaches to knowledge assessment, computation, and explanation under scarce resources =-=[15, 14]-=-. That is, we have sought to apply the principles of normative rationality to reason about the solution of a base-level decision problem. We focus our attention on real-time decision making. Resource-... |

1 |
Reasoning about inference tradeoffs in a world of bounded resources
- Horvitz
- 1986
(Show Context)
Citation Context ...ing such resource insufficiencies. We have been exploring resource constraints and the feasibility of normative approaches to knowledge assessment, computation, and explanation under scarce resources =-=[15, 14]-=-. That is, we have sought to apply the principles of normative rationality to reason about the solution of a base-level decision problem. We focus our attention on real-time decision making. Resource-... |

1 |
Reconciliation of decision analyses
- Lindley
- 1986
(Show Context)
Citation Context ...ion analysis or continuing to refine a decision. Matheson [24] explored the value of spending additional effort to analyze a bidding problem proposed by Howard [20]. Watson and Brown [37] and Lindley =-=[22]-=- describe issues surrounding the application of a preliminary decision analysis to assist in the decision as to whether an individual should embark on a more costly decision analysis of the base probl... |

1 |
Reasoning about inference tradeos in a world of bounded resources
- Horvitz
- 1986
(Show Context)
Citation Context ...dling such resource insuciencies. We have been exploring resource constraints and the feasibility of normative approaches to knowledge assessment, computation, and explanation under scarce resources =-=[15, 14]-=-. That is, we have sought to apply the principles of normative rationality to reason about the solution of a base-level decision problem. We focus our attention on real-time decision making. Resource-... |