In recent years, interface agents have been developed to assist users with various tasks. Some systems employ machine learning techniques to allow the agent to adapt to the user's changing requirements. With the increase in the volume of data on the Internet, agents have emerged which are able to monitor and learn from their users to identify topics of interest. One such agent, described here, has been developed to filter mail messages. We examine the issues involved in constructing an autonomous interface agent which employs a learning component, and explore the use of two different learning techniques in this context. Submitted to Applied Artificial Intelligence Journal. October 26, 1 INTRODUCTION 1 1 Introduction Agents were once seen as anthropomorphic entities which would assist users with daily tasks. They could be used, for example, to locate information of interest to their user (Kay 1984). Ten years later, many definitions of agents have been proposed. The basic concept of ...

A knowledge acquisition tool should provide a user with maximum guidance in extending and debugging a knowledge base, by preventing inconsistencies and knowledge gaps that may arise inadvertently. Most current acquisition tools are not very exible in that they are built for a predetermined inference structure or problem-solving mechanism, and the guidance they provide is specific to that inference structure and hard-coded by their designer. This paper focuses on expect, a reflective architecture that supports knowledge acquisition based on an explicit analysis of the structure of a knowledge-based system, rather than on a fixed set of acquisition guidelines. expect's problem solver is tightly integrated with loom, a state-of-the-art knowledge representation system. Domain facts and goals are represented declaratively, and the problem solver keeps records of their functionality within the task domain. When the user corrects the system's knowledge, expect tracks any possible implications of this change in the overall system and cooperates with the user to correct any potential problems that may arise. The key to the exibility of this knowledge acquisition tool is that it adapts its guidance as the knowledge bases evolve in response to changes

Any learning process can be viewed as a self-modification of the leaxnefs current knowledge tArough an. interaction with some information source. Such knowledge modification is guided by the learner's deshe to achieve a certain outcome, and can engage any kind of inference. The type of inference involved depends on he input information, the current (background) knowledge and the learneFs task ax hand. Based on such a view of learning, several fundamental concepts are analized and clarified, in paxticular, analytic and synthetic learning, derivm:ional and hypothetical explanation, constnictive induction, abduction, abstraction and deductive generalization. It is shown that inductive generalization and abduction can be viewed as two basic forms of general induction, and that abstraction and deductive generalization axe two related forms of constructive deduction. Using this conceptual framework, a methodology for multistrategy task-adaptive learning (MTL) is outlined, in which learning strategies axe combined dynamically, depending on the current learning situation. Speccally, an MTL learner anaLizes a "wiad" relationship among the input information, the background knowledge and the learning task, and on that basis determines which strategy, or. a combination thereof, is most appropriate at a given learning step. To implement the MTL methodology, a new knowledge representation is proposed, based on the parametric association rules (PARs). Basic ideas of MTL are illustrated by means of the well-known "cup" example, through which is shown how an MTL learner can employ, depending the above mad relationship, emprical learning, constructive inductive generalization, abduction, explanation-based learning and absuaction.

In a Bayesian framework, we give a principled account of how domainspecific prior knowledge such as imperfect analytic domain theories can be optimally incorporated into networks of locally-tuned units: by choosing a specific architecture and by applying a specific training regimen. Our method proved successful in overcoming the data deficiency problem in a large-scale application to devise a neural control for a hot line rolling mill. It achieves in this application significantly higher accuracy than optimally-tuned standard algorithms such as sigmoidal backpropagation, and outperforms the state-of-the-art solution. 1 INTRODUCTION Learning in connectionist networks typically requires many training examples and relies more or less explicitly on some kind of syntactic preference bias such as "minimal architecture" (Rumelhart, 1988; Le Cun et al., 1990; Weigend, 1991; inter alia) or a smoothness constraint operator (Poggio et al., 1990), but does not make use of explicit representations...

As knowledge-based systems are addressing increasingly complex domains, their roles are shifting from classical expert systems to interactive assistants. To develop and maintain such systems, an integration of thorough knowledge acquisition procedures and sustained learning from experience is called for. A knowledge level modeling perspective has shown to be useful for analyzing the various types of knowledge related to a particular domain and set of tasks, and for constructing the models of knowledge contents needed in an intelligent system. To be able to meet the requirements of future systems with respect to robust competence and adaptive learning behavior, particularly in open and weak theory domains, a stronger emphasis should be put on the combined utilization of casespecific and general domain knowledge. In this chapter we present a framework for integrating KA and ML methods within a total knowledge modeling cycle, favoring an iterative rather than a top down approac...

This paper describes our method for improving the comprehensibility, accuracy, and generality of reactive plans. A reactive plan is a set of reactive rules. Our method involves two phases: (1) formulate explanations of execution traces, and then (2) generate new reactive rules from the explanations. Since the explanation phase has been previously described, the primary focus of this paper is the rule generation phase. This latter phase consists of taking a subset of the explanations and using these explanations to generate a set of new reactive rules to add to the original set. The particular subset of the explanations that is chosen yields rules that provide new domain knowledge for handling knowledge gaps in the original rule set. The original rule set, in a complimentary manner, provides expertise to fill the gaps where the domain knowledge provided by the new rules is incomplete.

This paper is concerned with the task of incrementally acquiring and refining the knowledge and algorithms of a knowledge-based system in order to improve its performance over time. In particular, we present the design of DE-KART, a tool whose goal is to provide increasing levels of assistance in acquiring and refining indexing and retrieval knowledge for a knowledge-based retrieval system. DE-KART starts with knowledge that has been entered manually, and increases its level of assistance in acquiring and refining that knowledge, both in terms of the increased level of automation in interacting with users, and in terms of the increased generality of the knowledge. DE-KART is at the intersection of machine learning and knowledge acquisition: it is a first step towards a system which moves along a continuum from interactive knowledge acquisition to increasingly automated machine learning as it acquires more knowledge and experience.

This paper provides a taxonomy of constructive induction problems and reports on an empirical comparison of several constructive induction methods. In this paper a representation space is said to be poorly suited for learning because of three types of problems: 1) inappropriate attributes or attribute values, 2) incomplete attribute values, attribute sets or examples and 3) incorrect attributes or examples. Most current constructive induction methods are designed to correct one of these types (or sub-types) of problems which limits the types of problems for which this method is effective. In order to build a more general multistrategy method of constructive induction an understanding of when some methods for constructive induction are useful and when they fail is important. Five methods of constructive induction are evaluated: DCI attribute construction (AQDCI) , HCI attribute construction (AQ-HCI(ADD)), HCI attribute removal (AQ-HCI(REMOVE)), HCI construction and removal (AQ-HCI), and...

This paper describes a method for improving the comprehensibility, accuracy, and generality of reactive plans. A reactive plan is a set of reactive rules. Our method involves two phases: (1) formulate explanations of execution traces, and (2) generate new reactive rules from the explanations. The explanation phase involves translating the execution trace of a reactive planner into an abstract language, and then using Explanation Based Learning to identify general strategies within the abstract trace. The rule generation phase consists of taking a subset of the explanations and using these explanations to generate a set of new reactive rules to add to the original set for the purpose of performance improvement. 1. Introduction Reactive planning has proven to be a highly effective approach to planning (e.g., [10]). We have developed an enhancer for reactive plans that satisfies two goals. The first goal is to facilitate human understanding of plans generated by a particular class of rea...

Any learning process can be viewed as a self-modification of the leamer's current knowledge through an interaction with some information source. Such knowledge modification s graded by the learner s destre to achieve a certain outcome, and can engage any kind of inference. The typ0 of inference involved depends on the input information, the current (background) knowledge and the learne's task,.at h, and: Based on such a view of learning, several fundamental concepts are ananzeu ano clarified, in particular, analytic and synthetic learning, derivational and hypothetical explanation, constructive induction, abduction, abstraction and deductive generalization. It is shown that inductive generalization and abduction can be viewed as two basic forms of general induction, and that abstraction and deductive generalization are two related forms of constructive deduction. Using this conceptual framework, a methodology for multistrategy task-adaptive learning (MTL) is outlined, in which learning strategies are combined dynamically, depending on the current learning situation. Specifically, an MTL learner anali?es a "triad" relationship among the input information, the background knowledge and the learning task, and on that basis determines which strategy, or a combination thereof, is most appropriate at a given learning step. To implement the MTL methodology, a new knowledge representation is proposed, based on the parametric association rules (PARs). Basic ideas of MTL are illustrated by means of the well-known "cup" example, through which is shown how an MTL leamer can employ, depending on the above triad relationship, emprical learning, constructive inductive generalization, abduction, explanation-based learning and abstraction.