Hybrid learning methods use theoretical knowledge of a domain and a set of classified examples to develop a method for accurately classifying examples not seen during training. The challenge of hybrid learning systems is to use the information provided by one source of information to offset information missing from the other source. By so doing, a hybrid learning system should learn more effectively than systems that use only one of the information sources. KBANN(Knowledge-Based Artificial Neural Networks) is a hybrid learning system built on top of connectionist learning techniques. It maps problem-specific "domain theories", represented in propositional logic, into neural networks and then refines this reformulated knowledge using backpropagation. KBANN is evaluated by extensive empirical tests on two problems from molecular biology. Among other results, these tests show that the networks created by KBANN generalize better than a wide variety of learning systems, as well as several t...

This paper describes a successful approach to concept learning for heuristic classification. Almost all current programs for this task create or use explicit, abstract generalizations. These programs are largely ineffective for domains with weak or intractable theories. An exemplar-based approach is suitable for domains with inadequate theories but raises two additional problems: determining similarity and indexing exemplars. Our approach extends the exemplar-based approach with solutions to these problems. An implementation of our approach, called Protos, has been applied to the domain of clinical audiology. After reasonable training, Protos achieved a competence level equaling that of human experts and far surpassing that of other machine learning programs. Additionally, an "ablation study" has identified the aspects of Protos that are primarily responsible for its success. 1 Introduction This paper describes a successful approach to the task of concept learning for heuristic clas...

This paper presents a view of planning as a task supported by a dynamic memory. This view attempts to integrate models of memory, learning and planning into a single system that learns about planning by creating new plans and analyzing how they interact with the world. We call this view of planning Case-Based Planning. A case-based planner makes use of its own past experience in developing new plans. It relies on its memory of observed effects, rather than a set of causal rules, to create and modify new plans. Memories of past successes are accessed and modified to create new plans. Memories of past failures are used to warn the planner of impending problems, and memories of past repairs are called upon to tell the planner how to how to deal with them. This view of planning from experience supports and is supported by a learning system that incorporates new experiences into the planner's episodic memory. This learning algorithm gains from the planner's failures as well as its successe...

Theories of categorization have typically focused on the internal structure of categories. This paper is concerned with the external structure of categories. In particular , it is suggested that many categories specify the relational role that is played by category members. To support this claim, the paper distinguishes between traditional feature-based categories, relational categories (which specify a relational structure) and role-governed categories (which specify that an item plays a particular role within a relational structure). After discussing the relationship among these types of categories, the implications of this view for the study of category learning and category use are discussed.

ion and Categorization Eric J. Horvitz # and Adrian C. Klein Palo Alto Laboratory Rockwell International Science Center 444 High Street Palo Alto, CA 94301 Abstract We take a utility-based approach to categorization. We construct generalizations about events and actions by considering losses associated with failing to distinguish among detailed distinctions in a decision model. The utility-based methods transform detailed states of the world into more abstract categories comprised of disjunctions of the states. We show how we can cluster distinctions into groups of distinctions at progressively higher levels of abstraction, and describe rules for decision making with the abstractions. The techniques introduce a utility-based perspective on the nature of concepts, and provide a means of simplifying decision models used in automated reasoning systems. We demonstrate the techniques by describing the capabilities and output of TUBA, a program for utility-based abstraction. 1 INTRODUCTION...

Reasoning from prior cases or abstractions requires that a system identify relevant similarities between the current situation and objects represented in memory. Often, relevance depends upon abstract, thematic, costly-to-infer properties of the situation. Because of the cost of inference, a case retrieval system needs to learn which descriptions are worth inferring, and how costly that inference will be. This paper outlines the properties that make an abstract thematic feature valuable to a case-based reasoner, and recasts the problem of case retrieval into a framework under which a system can explicitly and dynamically reason about the cost of acquiring features relative to their information value. 1 Retrieval, description, and learning For a case-based reasoner to make effective use of recalled prior experiences, it must be able to judge which of its cases are applicable to the current situation. This problem is not new nor is it unique to case-based reasoning: any system that re...

This paper is a treatment of the problem of concept acquisition by autonomous agents, primarily from an AI point of view. However, as this problem is not very well studied in AI and as humans are indeed a kind of autonomous agent, the problem is also studied from a psychological point of view to see if the research in human concept acquisition can be of any help when designing artificial agents. However, the acquisition cannot be studied in isolation since it is dependent on more fundamental aspects of concepts. Consequently, these are studied as well. Thus, this paper will give a review of some of the research done in cognitive psychology and AI (and to some extent philosophy) on different aspects of concepts. Some proposals for how central problems should be attacked and some pointers for further research are also presented. 1 Introduction In order to pursue goals and to plan future actions efficiently, an intelligent system (human or artificial) must be able to classify and reason ...

This paper describes Zenith, a discovery system that performs constructive induction. The system is able to generate and extend new features for concept learning using agenda-based heuristic search. The search is guided by feature worth (a composite measure of discriminability and cost). Zenith is distinguished from existing constructive induction systems by its interaction with a performance system, and its ability to extend its knowledge base by creating new domain classes. Zenith is able to discover known useful features for the Othello board game. Feature Discovery for Inductive Concept Learning 1 1 Introduction One of the central concerns of machine learning is that of inductive concept learning from examples, in which a system is given a set of examples and produces a characterization of them. Many induction algorithms have been devised that are able to inductively generalize in different formalisms, using learning rules appropriate for that formalism. For example, decision t...

The TeacherBridge (Basic Resources for Integrated Distributed Group Environments) motivates collaboration and supports online tools for teachers ’ professional resource management by providing a socio-technical infrastructure for community networks. While this project focuses initially on local science and mathematics teachers, resources developed are designed to be globally diffused to a wide variety of online communities. This paper discusses the characteristics of online communities of educators and how home networking technology and knowledge management systems can support collaboration and knowledge sharing in such communities. We study examples of existing and well-known online communities of educators and introduce our own system, TeacherBridge, which supports teacher professional development by supporting peer-based collaboration and community. We also analyze and evaluate the characteristics of TeacherBridge with activity theory [12] and minimalism [37]. Activity theory provides an analytical framework for how TeacherBridge can be used as a socio-technical infrastructure for online communities of educators; minimalism guides the development of successful online communities that are easily accessible and facilitate teacher participation in knowledge sharing activities

Introduction Cognitive scientists have been, for years, searching for essential ingredients of intelligence. Although this issue may not be satisfactorily resolved for quite some time, two abilities are clearly central to intelligent behaviour. One is the ability to acquire knowledge or skill through experience; that is, the ability to learn. The second is the ability to apply the knowledge or skill possessed to solve new problems; that is, the ability to reason. The new problems may concern actual events in the real world: for example, when one has to react to a new external stimulus; or may be imaginary, for instance, when one creates them for planning purposes. A precondition for the above abilities is the capability to represent diverse forms of knowledge. As our knowledge is built of individual concepts, to represent knowledge one needs to represent concepts. Consequently, understanding how concepts are represented is a fundamental problem underlying all efforts in t