Showing papers on "Adaptive reasoning published in 1990"
Proceedings Article•
01 Jan 1990TL;DR: Two systems that attempt to combined case-based methods with model-based - explanation-ba sed - approaches are described and some of their weaknesses are identified, and an improved approach to integration of case- based and model- based methods - called CREEK 1 - is suggested.
Abstract: reasoning (CBR) a problem is solved by matching the problem description to a previously solved case, using the past solution in solving the new problem. A case-based reasoner learns after each problem solving session by retaining relevant information from a problem just solved, making the new experience available for future problem solving. Crucial steps in a CBR process include finding a good match to a new problem, adapting a previous solution to successfully solve the new problem, and deciding how to index and store a new case for later effective retrieval. Previous CBR systems relied on syntactic rather than semantic or pragmatic criteria in performing these steps. A comprehensive model of general domain knowledge is needed in order to match cases based on their meaning contents. This has lead to systems that attempt to combined case-based methods with model-based - explanation-ba sed - approaches. Two systems representative of this research, PROTOS and CASEY, are briefly described. The systems are discussed with respect to what type of general knowledge they contain, and the degree of model-based support for the case-based reasoning and learning processes they exhibit. Some of their weaknesses are identified, and an improved approach to integration of case-based and model-based methods - called CREEK 1 - is suggested.
66 citations
TL;DR: Four reasoning algorithms for the G-net model are proposed: inheritance reasoning and recognition reasoning for semantic knowledge, event-driven reasoning for dynamic knowledge, and control table reasoning for coordination and control in a mixed-type knowledge hierarchy.
Abstract: The G-net model for G-type knowledge representation is introduced. It is capable of modeling both static semantic knowledge and dynamic control knowledge, combining them into a loosely coupled, mixed-type knowledge hierarchy. Four reasoning algorithms for the G-net model are proposed: inheritance reasoning and recognition reasoning for semantic knowledge, event-driven reasoning for dynamic knowledge, and control table reasoning for coordination and control in a mixed-type knowledge hierarchy. Based on the knowledge-table representation, the G-net model expresses the constraints and relationships among knowledge objects explicitly so that reasoning algorithms can be implemented efficiently. Applications to information systems prototyping are discussed. >
62 citations
Proceedings Article•
29 Jul 1990TL;DR: DARES is presented, a distributed reasoning system in which agents have the ability to focus their attention on selective information interchange to facilitate cooperative problem solving and suggests that an effective cooperation strategy which is largely independent of initial knowledge distribution is developed.
Abstract: In many domains of interest to distributed artificial intelligence, the problem solving environment may be viewed as a collection of loosely coupled intelligent agents, each of which reasons based on its own incomplete knowledge of the state of the world No agent has sufficient knowledge to solve the problem at hand so that coordinated cooperative problem solving is required to satisfy system goals In this paper, we present DARES, a distributed reasoning system in which agents have the ability to focus their attention on selective information interchange to facilitate cooperative problem solving The experimental results we present demonstrate that agents in a loosely coupled network of problem solvers can work semi-independently, yet focus their attention with the aid of relatively simple heuristics when cooperation is appropriate These results suggest that we have developed an effective cooperation strategy which is largely independent of initial knowledge distribution
49 citations
01 Jan 1990
TL;DR: Five capabilities that are necessary to effect default reasoning are identified and the major characteristics for systems that handle incomplete and uncertain information as well as other types of imperfect information are established.
Abstract: Two general approaches to reasoning with imperfect information are discussed: nonmonotonic reasoning and a calculus of uncertainty. Default reasoning is posed as an approach that is potentially capable of integrating many facets of these two approaches. Practical requirements for default reasoning are then established. This is done by identifying a number of cases that involve incomplete and uncertain information and showing how they can be addressed by default reasoning. Parametric and symbolic reasoning are differentiated, and it is shown that both types are necessary. This distinction is important, as most approaches tend to neglect either the parametric or the symbolic aspect of default reasoning, thereby restricting its use to one of the two approaches discussed above. Five capabilities that are necessary to effect default reasoning are identified. The major characteristics for systems that handle incomplete and uncertain information as well as other types of imperfect information are established. >
26 citations
10 citations
TL;DR: This article focuses on the relationship between two approaches to the automation of reasoning: the paradigm based on the clause language and that based on natural deduction.
Abstract: This article is the fourteenth of a series of articles discussing various open research problems in automated reasoning. Here we focus on the relationship between two approaches to the automation of reasoning: the paradigm based on the clause language and that based on natural deduction. The problem proposed for research asks one to find a mapping between the two approaches such that reasoning programs based on either approach perform identically on a specific assignment. For evaluating a proposed solution to this research problem, we include suggestions concerning possible test problems.
8 citations
TL;DR: In this article, the authors suggest that the concepts and formalisms developed within this branch of artificial intelligence can be usefully applied to the study of political institutions and political behavior, and argue, critically, that if CBR is to be applied to politics, it must be generalized to accommodate multiple agents who act repeatedly in multiple tasks.
Abstract: Case-based reasoning (CBR) is an active area of research within artificial intelligence that emphasizes the function of memory in problem-solving. It proceeds from the premise that people cope with new problems or situations by reusing the strategies that have proved effective in similar situations in the past. Rather than deduction or the application of rules, the basic inferential process is that of recognition. Because the reasoning characteristic of politics is largely of this type, we suggest that the concepts and formalisms developed within this branch of artificial intelligence can be usefully applied to the study of political institutions and political behavior. But we argue, critically, that if CBR is to be applied to politics, it must be generalized to accommodate multiple agents who act repeatedly in multiple tasks. We outline what these extensions to CBR would involve, using the history of liability in tort law as a worked example.
8 citations
7 citations
TL;DR: The authors express the view that algorithms and heuristics may be used to help improve professional problem solving abilities provided that they are appropriately contextualised within the relevant discipline.
Abstract: The authors express the view that algorithms and heuristics may be used to help improve professional problem solving abilities provided that they are appropriately contextualised within the relevant discipline. The first part of this paper presents a basic algorithm applied to the engineering discipline. A similar algorithm applicable in the solving of legal problems is presented later in the paper. Problem complexity and related variables are discussed as are ways of teaching problem solving and integrating problem solving into the curriculum. Typical problem solving exercises are described.
06 Nov 1990
TL;DR: A novel logic named fuzzy computational reasoning is presented, and two concrete reasoning methods based on the logic, data- driven computational reasoning and goal-driven computational reasoning are proposed.
Abstract: A novel logic named fuzzy computational reasoning is presented, and two concrete reasoning methods based on the logic, data-driven computational reasoning and goal-driven computational reasoning are proposed. In these reasoning procedures, reasoning is computing. In addition, the author presents the concept of a fuzzy neural network in which a neuron is represented by an inference rule and a neural network is represented by a set of inference rules. The behavior models of the neural networks are based on the computational reasoning. Finally, a descriptive language to describe the fuzzy neural networks is designed. >
05 Dec 1990
TL;DR: The author presents the basic concepts of qualitative reasoning (QR) and then looks at what one needs to do to use it for control and suggests that, in order to reason about control within this framework, the framework needs to include feedback and learning capabilities and to relax some of its assumptions.
Abstract: The author presents the basic concepts of qualitative reasoning (QR) and then looks at what one needs to do to use it for control. In QR, physical problems are treated as a general domain and the ideas of forces and other physical phenomena are incorporated into the problem solving structure itself. It is suggested that, in order to reason about control within this framework, the framework needs to include feedback and learning capabilities and to relax some of its assumptions. >
TL;DR: Basic, theoretical backgrounds for dynamic backward reasoning systems can provide basic concepts to be incorporated into a specialized planning or expert system for support of knowledge processing in realistic domains.
01 Oct 1990
TL;DR: Under the sponsorship of DARPA, a generic CBR shell is developed that was evaluated in the domain of NACA airfoils and retrieved 9 out of 17 selected by the expert, after modifying the similarity algorithms.
Abstract: Case-Based Reasoning (CBR) is a methodology for employing imprecise data and uncertain information in the development of solutions to fuzzy real world problems. It is seen as an alternative to rule-based systems, which may fail under these conditions. Under the sponsorship of DARPA, we have developed a generic CBR shell.The system was evaluated in the domain of NACA airfoils. A subject matter expert was asked to select airfoils (cases) which were similar to target airfoils. He then defined attributes and weights by which he had judged this similarity. These parameters were then used by PROSPER in a retrieval of airfoils similar to the same targets. From a case base of 98 airfoils, PROSPER retrieved 9 out of 17 selected by the expert. After modifying the similarity algorithms, PROSPER retrieved 11 out of 17.
01 Jan 1990
TL;DR: In this paper, the authors present a survey of the state-of-the-art tools and techniques used in the design of the HOGA algorithm and its application in the context of artificial intelligence applications.
Abstract: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3
01 Jan 1990
TL;DR: This study investigates a number of approaches including Bayesian Probability, Mycin Certainty Factors, Dempster-Shafer Theory of Evidence, Fuzzy Set Theory, Possibility Theory and non monotonic logic for handling uncertainty in problem solving situations.
Abstract: Much of the research done in Artificial Intelligence involves investigating and developing methods of incorporating uncertainty reasoning and representation into expert systems. Several methods have been proposed and attempted for handling uncertainty in problem solving situations. The theories range from numerical approaches based on strict probabilistic reasoning to non-numeric approaches based on logical reasoning. This study investigates a number of these approaches including Bayesian Probability, Mycin Certainty Factors, Dempster-Shafer Theory of Evidence, Fuzzy Set Theory, Possibility Theory and non monotonic logic. Each of these theories and their underlying formalisms are explored by means of examples. The discussion concentrates on a comparison of the different approaches, noting the type of uncertainty that they best represent. CR Categories 1.2.1 Applications and Expert Systems Medicine and Science 1.2.4 Artificial Intelligence Knowledge Representation Formalisms and Methods
06 Nov 1990
TL;DR: An effective high-level reasoning method, i.e. the multidimensional argument method, is provided which can improve the accuracy of the whole cooperative system's results by using various kinds of information, such as mutual-evaluation information, self- evaluation, external knowledge and information, and dependence value.
Abstract: The authors discuss high-level reasoning in a cooperative knowledge system and provide an effective high-level reasoning method, i.e. the multidimensional argument method. Based on the results of low-level reasoning this method of high-level reasoning can improve the accuracy of the whole cooperative system's results by using various kinds of information, such as mutual-evaluation information, self-evaluation, external knowledge and information, and dependence value. The authors also describe the development of the argument system PAT-1 which is a subsystem of DAI-tools HEDMUKCS. >
Journal Article•
TL;DR: A method of inference is introduced which combines approximate reasoning and default logic, and the procedure of transforming monotonic reasoning into default reasoning is given.
Abstract: Fuzzy set systems can be used to solve the problem with uncertain knowledge, and default logic can be used to solve the problem with incomplete knowledge, in some sense. In this paper, based on interval-valued fuzzy sets we introduce a method of inference which combines approximate reasoning and default logic, and give the procedure of transforming monotonic reasoning into default reasoning.
01 Jan 1990
TL;DR: The model proposed in this paper belongs to a new class of data models that are characterized by the ability to capture both data semantics and knowledge and incorporates features taken from the knowledge based systems area such as inferential capabilities.
Abstract: The model proposed in this paper belongs to a new class of data models that are characterized by the ability to capture both data semantics (description of entities ) and knowledge (inferential and operational capabilities). Our approach attempts to extend the capabilities of semantic data models by incorporating features taken from the knowledge based systems area such as inferential capabilities. Its main contribution includes a mechanism that allows abstract knowledge typing by handling all concepts: heuristics, properties, relationships and constraints as objects. It also proposes an approach to reasoning process modelisation and presents concepts and mechanisms for problem solving simulation.
TL;DR: This article focuses on finding criteria for correctly choosing between using logic programming and a more general automated reasoning approach to attack a given assignment.
Abstract: This article is the thirteenth of a series of articles discussing various open research problems in automated reasoning. Here we focus on finding criteria for correctly choosing between using logic programming and a more general automated reasoning approach to attack a given assignment. The problem proposed for research asks one to find criteria that classify problems as solvable with a well-focused algorithm or as requiring a more general search for new information. We include suggestions for evaluating a proposed solution to this research problem.