Kaiserslautern - Fachbereich Informatik
Refine
Year of publication
- 1996 (32) (remove)
Language
- English (32) (remove)
Has Fulltext
- yes (32)
Keywords
- AG-RESY (4)
- COMOKIT (3)
- PARO (3)
- Case-Based Reasoning (2)
- CoMo-Kit (2)
- Fallbasiertes Planen (2)
- SKALP (2)
- case-based planning (2)
- Abstraction (1)
- General Knowledge (1)
Faculty / Organisational entity
Evolving Combinators
(1996)
One of the many abilities that distinguish a mathematician from an auto-mated deduction system is to be able to offer appropriate expressions based onintuition and experience that are substituted for existentially quantified variablesso as to simplify the problem at hand substantially. We propose to simulate thisability with a technique called genetic programming for use in automated deduc-tion. We apply this approach to problems of combinatory logic. Our experimen-tal results show that the approach is viable and actually produces very promisingresults. A comparison with the renowned theorem prover Otter underlines theachievements.This work was supported by the Deutsche Forschungsgemeinschaft (DFG).
We present a similarity criterion based on feature weighting. Feature weights are recomputed dynamically according to the performance of cases during problem solving episodes. We will also present a novel algorithm to analyze and explain the performance of the retrieved cases and to determine the features whose weights need to be recomputed. We will perform experiments and show that the integration in a feature weighting model of our similarity criterion with our analysis algorithm improves the adaptability of the retrieved cases by converging to best weights for the features over a period of multiple problem solving episodes.
We present a concept for an automated theorem prover that employs a searchcontrol based on ideas from several areas of artificial intelligence (AI). The combi-nation of case-based reasoning, several similarity concepts, a cooperation conceptof distributed AI and reactive planning enables a system using our concept tolearn form previous successful proof attempts. In a kind of bootstrapping processeasy problems are used to solve more and more complicated ones.We provide case studies from two domains of interest in pure equationaltheorem proving taken from the TPTP library. These case studies show thatan instantiation of our architecture achieves a high grade of automation andoutperforms state-of-the-art conventional theorem provers.
EADOCS (Expert Assisted Design of Composite Structures) is the implementation of a multi-level approach to conceptual design. Constraint-, case- and rule-based reasoning techniques are applied in different design phases to assemble and adapt designs at increasing levels of detail. This paper describes a strategic approach to decomposition, formulation of target design problems, and incremental retrieval and adaptation. Design problems considered, cannot be decomposed dynamically into tractable subproblems. Design cases are retrieved for requirements and preferences on both functionality and the solution. Cases are adapted in three phases: adaptation, modification and optimisation.
We investigate the usage of so-called inference rights. We point out the prob-lems arising from the inflexibility of existing approaches to heuristically controlthe search of automated deduction systems, and we propose the application ofinference rights that are well-suited for controlling the search more flexibly. More-over, inference rights allow for a mechanism of "partial forgetting" of facts thatis not realizable in the most controlling aproaches. We study theoretical founda-tions of inference rights as well as the integration of inference rights into alreadyexisting inference systems. Furthermore, we present possibilities to control suchmodified inference systems in order to gain efficiency. Finally, we report onexperimental results obtained in the area of condensed detachment.The author was supported by the Deutsche Forschungsgemeinschaft (DFG).
When problems are solved through reasoning from cases, the primary kind of knowledge is contained in the specific cases which are stored in the case base. However, in many situations additional background-knowledge is required to cope with the requirements of an application. We describe an approach to integrate such general knowledge into the reasoning process in a way that it complements the knowledge contained in the cases. This general knowledge itself is not sufficient to perform any kind of model-based problem solving, but it is required to interpret the available cases appropriately. Background knowledge is expressed by two different kinds of rules that both must be formalized by the knowledge engineer: Completion rules describe how to infer additional features out of known features of an old case or the current query case. Adaptation rules describe how an old case can be adapted to fit the current query. This paper shows how these kinds of rules can be integrated into an object-oriented case representation.
Planning for realistic problems in a static and deterministic environment with complete information faces exponential search spaces and, more often than not, should produce plans comprehensible for the user. This article introduces new planning strategies inspired by proof planning examples in order to tackle the search-space-problem and the structured-plan-problem. Island planning and refinement as well as subproblem refinement are integrated into a general planning framework and some exemplary control knowledge suitable for proof planning is given.
Software development organizations measure their real-world processes, products, and resources to achieve the goal of improving their practices. Accurate and useful measurement relies on explicit models of the real-world processes, products, and resources. These explicit models assist with planning measurement, interpreting data, and assisting developers with their work. However, little work has been done on the joint use of measurem(int and process technologies. We hypothesize that it is possible to integrate measurement and process technologies in a way that supports automation of measurement-based feedback. Automated support for measurementbased feedback means that software developers and maintainers are provided with on-line, detailed information about their work. This type of automated support is expected to help software professionals gain intellectual control over their software projects. The dissertation offers three major contributions. First, an integrated measurement and
process modeling framework was constructed. This framework establishes the necessary foundation for integrating measurement and process technologies in a way that will permit automation. Second, a process-centered software engineering environment was developed to support measurement-based feedback. This system provides personnel with information about the tasks expected of them based on an integrated set of measurement and process views. Third, a set of assumptions and requirements about that system were examined in a controlled experiment. The experiment compared the use of different levels of automation to evaluate the acceptance and effectiveness of measurement-based feedback.
This paper addresses the role of abstraction in case-based reasoning. We develop a general framework for reusing cases at several levels of abstraction, which is particularly suited for describing and analyzing existing and designing new approaches of this kind. We show that in synthetic tasks (e.g. configuration, design, and planning), abstraction can be successfully used to improve the efficiency of similarity assessment, retrieval, and adaptation. Furthermore, a case-based planning system, called Paris, is described and analyzed in detail using this framework. An empirical study done with Paris demonstrates significant advantages concerning retrieval and adaptation efficiency as well as flexibility of adaptation. Finally, we show how other approaches from the literature can be classified according to the developed framework.
We present a novel approach to classification, based on a tight coupling of instancebased learning and a genetic algorithm. In contrast to the usual instance-based learning setting, we do not rely on (parts of) the given training set as the basis of a nearestneighbor classifier, but we try to employ artificially generated instances as concept prototypes. The extremely hard problem of finding an appropriate set of concept prototypes is tackled by a genetic search procedure with the classification accuracy on the given training set as evaluation criterion for the genetic fitness measure. Experiments with artificial datasets show that - due to the ability to find concise and accurate concept descriptions that contain few, but typical instances - this classification approach is considerably robust against noise, untypical training instances and irrelevant attributes. These favorable (theoretical) properties are corroborated using a number of hard real-world classification problems.