Abstract. Storing and using specific instances improves the performance of several supervised learning algorithms. These include algorithms that learn decision trees, classification rules, and distributed networks. However, no investigation has analyzed algorithms that use only specific instances to solve incremental learning tasks. In this paper, we describe a framework and methodology, called instance-based learning, that generates classification predictions using only specific instances. Instance-based learning algorithms do not maintain a set of abstractions derived from specific instances. This approach extends the nearest neighbor algorithm, which has large storage requirements. We describe how storage requirements can be significantly reduced with, at most, minor sacrifices in learning rate and classification accuracy. While the storage-reducing algorithm performs well on several realworld databases, its performance degrades rapidly with the level of attribute noise in training instances. Therefore, we extended it with a significance test to distinguish noisy instances. This extended algorithm's performance degrades gracefully with increasing noise levels and compares favorably with a noise-tolerant decision tree algorithm.

This chapter provides an introduction to case-based reasoning, discusses motivations for CBR, and describes the central steps in the CBR process. It examines the relationship of CBR to other approaches and discusses major research areas, open issues, and promising opportunities for CBR. It surveys and relates numerous approaches within CBR and provides more than 150 references to international CBR research.

CATO is an intelligent learning environment designed to help beginning law students learn basic skills of making arguments with cases. Using CATO, students practice tasks of induction and analogical argumentation. They practice testing theories against a body of cases and making written arguments about a problem, comparing and contrasting it to past cases. CATO’s model addresses arguments in which two opponents analogize a problem to favorable cases, distinguish unfavorable cases, assess the significance of similarities and differences between cases in light of normative knowledge about the domain, and use that knowledge to organize multi-case arguments. CATO communicates the model to students by presenting dynamically-generated argumentation examples and by reifying argument structure based on the model. CATO also provides a case database and tools based on the model that help make students ’ tasks more manageable. CATO was evaluated in the context of an actual legal writing course, in a study involving 30 first-year law students. We found that instruction with CATO leads to statistically significant improvement in students ’ basic argumentation skills, comparable

Practical scheduling problems generally require allocation of resources in the presence of a large, diverse and typically conflicting set of constraints and optimization criteria. The ill-structuredness of both the solution space and the desired objectives make scheduling problems difficult to formalize. This paper describes a case-based learning method for acquiring context-dependent user optimization preferences and tradeoffs and using them to incrementally improve schedule quality in predictive scheduling and reactive schedule management in response to unexpected execution events. The approach, implemented in the CABINS system, uses acquired user preferences to dynamically modify search control to guide schedule improvement. During iterative repair, cases are exploited for: (1) repair action selection, (2) evaluation of intermediate repair results and (3) recovery from revision failures. The method allows the system to dynamically switch between repair heuristic actions, each of whi...

Patdex is an expert system which carries out case-based reasoning for the fault diagnosis of complex machines. It is integrated in the Moltke workbench for technical diagnosis, which was developed at the university of Kaiserslautern over the past years, Moltke contains other parts as well, in particular a model-based approach; in Patdex where essentially the heuristic features are located. The use of cases also plays an important role for knowledge acquisition. In this paper we describe Patdex from a principal point of view and embed its main concepts into a theoretical framework 1 General Considerations Patdex 1 is an expert system which carries out case-based reasoning for the fault diagnosis of complex machines. It is integrated in the Moltke workbench 2 for technical diagnosis, which was developed at the university of Kaiserslautern over the past years (cf. e.g. [4, 5, 23]), Moltke contains other parts as well (cf. e.g. [16]), in particular a model-based approach (cf. [21, ...

. Case-based problem solving can be significantly improved by applying domain knowledge (in opposition to problem solving knowledge) , which can be acquired with reasonable effort, to derive explanations of the correctness of a case. Such explanations, constructed on several levels of abstraction, can be employed as the basis for similarity assessment as well as for adaptation by solution refinement. The general approach for explanation-based similarity can be applied to different real world problem solving tasks such as diagnosis and planning in technical areas. This paper presents the general idea as well as the two specific, completely implemented realizations for a diagnosis and a planning task. 1 Introduction and Motivation The underlying principle of case-based reasoning is the idea to remember solutions to already known problems for their reuse during novel problem solving. The case which is most similar to the current problem is retrieved from a case base and its solution is m...

Planning the ongoing care of Alzheimer's Disease (AD) patients is a complex task, marked by cases that change over time, multiple perspectives, and ethical issues. Geriatric interdisciplinary teams of physicians, nurses and social workers currently plan this care without computer assistance. Although AD is incurable, interventions are planned to improve the quality of life for patients and their families. Much of the reasoning involved is case-based, as clinicians look to case histories to learn which interventions are effective, to document clinical ndings, and to train future health care professionals. There is great variability among AD patients, and within the same patient over time. AD is not yet well enough understood for universally effective treatments to be available. The case-based reasoning (CBR) research paradigm complements the medical research approach of finding treatments effective for all patients by matching patients to treatments that were effective for similar patients in the past. The Auguste Project is an effort to provide decision support for planning the ongoing care of AD patients, using CBR and other thought processes natural to members of geriatric interdisciplinary teams. System prototypes are used to explore the reasoning processes involved and to provide the forerunners of practical clinical tools. The first system prototype has just been completed. This prototype supports the decision to prescribe neuroleptic drugs to AD patients with behavioral problems. It uses CBR to determine if a neuroleptic drug should be prescribed and rule-based reasoning to select one of ve approved neuroleptic drugs for a patient. The first system prototype serves as proof of concept that CBR is useful for planning ongoing care for AD patients. Additional prototypes are planned to explore the research issues raised.

Much recent research on modeling memory processes has focused on identifying useful indices and retrieval strategies to support particular memory tasks. Another important question concerning memory processes, however, is how retrieval criteria are learned. This paper examines the issues involved in modeling the learning of memory search strategies. It discusses the general requirements for appropriate strategy learning and presents a model of memory search strategy learning applied to the problem of retrieving relevant information for adapting cases in case-based reasoning. It discusses an implementation of that model, and, based on the lessons learned from that implementation, points towards issues and directions in refining the model. Introduction Much recent AI research on memory focuses on analyzing the indices that are relevant to particular classes of retrieval problems (e.g., (Domeshek, 1992; Leake, 1992; Owens, 1991)). The problem of how memory search strategies can be learned...

In [9] we introduced a formal framework for constructing ordinal similarity measures, and suggested how this might also be applied to cardinal measures. In this paper we will place this approach in a more general framework, called similarity metrics. In this framework, ordinal similarity metrics (where comparison returns a boolean value) can be combined with cardinal metrics (returning a numeric value) and, indeed, with metrics returning values of other types, to produce new metrics.

In case-based reasoning systems, the case adaptation process is traditionally controlled by static libraries of hand-coded adaptation rules. This paper proposes a method for learning adaptation knowledge in the form of adaptation strategies of the type developed and hand-coded by Kass [90] . Adaptation strategies differ from standard adaptation rules in that they encode general memory search procedures for finding the information needed during case adaptation; this paper focuses on the issues involved in learning memory search procedures to form the basis of new adaptation strategies. It proposes a method that starts with a small library of abstract adaptation rules and uses introspective reasoning about the system's memory organization to generate the memory search plans needed to apply those rules. The search plans are then packaged with the original abstract rules to form new adaptation strategies for future use. This process allows a CBR system not only to learn about its domain, b...