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...

One of the major assumptions in Artificial Intelligence is that similar experiences can guide future reasoning, problem solving and learning; what we will call, the similarity assumption. The similarity assumption is used in problem solving and reasoning systems when target problems are dealt with by resorting to a previous situation with common conceptual features. In this article, we question this assumption in the context of case-based reasoning (CBR). In CBR, the similarity assumption plays a central role when new problems are solved, by retrieving similar cases and adapting their solutions. The success of any CBR system is contingent on the retrieval of a case that can be successfully reused to solve the target problem. We show that it is unwarranted to assume that the most similar case is also the most appropriate from a reuse perspective. We argue that similarity must be augmented by deeper, adaptation knowledge about whether a case can be easily modified to fit a target problem. We implement this idea in a new technique, called adaptation-guided retrieval (AGR), which provides a direct link between retrieval similarity and adaptation needs. This technique uses specially formulated adaptation knowledge, which, during retrieval, facilitates the computation of a precise measure of a case’s adaptation requirements. In closing, we assess the broader implications of AGR and argue that it is just one of a growing number of methods that seek to overcome the limitations of the traditional, similarity assumption in an effort to deliver more sophisticated and scaleable reasoning systems. Smyth & Keane 3 Adaptation-Guided Retrieval 1

Abstract not found

this paper, we describe a framework for goal-driven learning and its relationship to prior and current theories from each of these perspectives.

Creative conceptual change involves (a) the construction of new concepts and of coherent belief systems, or theories, relating these concepts, and (b) the modification and extrapolation of existing concepts and theories in novel situations. We discuss these and other types of conceptual change, and present computational models of constructive and extrapolative processes in creative conceptual change. The models have been implemented as computer programs in two very different task domains, autonomous robotic navigation and fictional story understanding. Contents 1 Introduction 1 2 Case studies in creative conceptual change 4 3 Constructive conceptual change 6 4 Technical details: The SINS system 9 4.1 Task: Autonomous robotic navigation : : : : : : : : : : : : : : : : : : : : : : : 9 4.2 Representation: Continuous prototypical cases : : : : : : : : : : : : : : : : : : 13 4.3 Process: Concept construction and modification : : : : : : : : : : : : : : : : : : 17 4.4 Discussion: Construc...

Most conceptual clustering systems rely solely on data to form concepts without supervision; the few that exploit causalities in the background knowledge do so only after the completion of a similarity-based learning phase. In this paper, we describe a multistrategy misconception discovery system, MMD, that utilizes data and theory in a more tightly coupled way. The integration of similarity- and causality-based learning in MMD is shown to be essential for the automatic construction of accurate and meaningful misconceptions that account for errors in novice behavior. 1

We present a framework for research on coherence of student conceptions in philosophy education. Commonsense conceptions of philosophical novices were studied. Students of a Finnish upper secondary school with no prior background in philosophy were asked to evaluate statements on conceptual issues in the domains of philosophy of mind, metaphysics and epistemology. The results were visualized with Kohonen self-organizing- maps (SOM), enabling us to identify clusters of students and questions with similar response patterns. The results are interpreted in terms of students ’ ontological commitments.

We present a connectionist model of children's developing reliance on object labels as opposed to superficial appearance when making inductive inferences. The model learns to infer a fact about an object based on the object's label (and not percept) even though that fact has never been previously associated with the label. The shift in reliance from perceptual to label information is found to depend on: (a) the presence of a pre-linguistic ability to categorize perceptual information, and (b) the greater variability of percepts than labels The model predicts that children will shift their inductive basis at different ages depending on the perceptual variability of the test categories. This prediction is discussed with respect to studies of children's induction and with particular reference to conflicting results reported in the literature concerning the onset of label use. Introduction This paper presents a connectionist model of the child's developing reliance on objec...

Knowledge discovery from unlabeled data comprises two main tasks: identification of "natural groups" and analysis of these groups in order to interpret their meaning. These tasks are accomplished by unsupervised and supervised learning, respectively, and correspond to the taxonomy and explanation phases of the discovery process described by Langley [9]. The efforts of Knowledge Discovery from Databases (KDD) research field has addressed these two processes into two main dimensions: (1) scaling up the learning algorithms to very large databases, and (2) improving the efficiency of the knowledge discovery process. In this paper we argue that the advances achieved in scaling up supervised and unsupervised learning algorithms allow us to combine these two processes in just one model, providing extensional (who belongs to each group) and intensional (what features best describe each group) descriptions of unlabeled data. To explore this idea we present an artificial neural network (ANN) architecture, using as building blocks two well-know models: the ART1 network, from the Adaptive Resonance Theory family of ANNs [4], and the Combinatorial Neural Model (CNM), proposed by Machado ([11] and [12])). Both models satisfy one important desiderata for data mining, learning in just one pass of the database. Moreover, CNM, the intensional part of the architecture, allows one to obtain rules directly from its structure. These rules represent the insights on the groups. The architecture can be extended to other supervised/unsupervised learning algorithms that comply with the same desiderata. ? Researcher at EMBRAPA --- Brazilian Enterprise for Agricultural Research and lecturer at Catholic University of Bras ilia (Supported by CAPES - Coordenacao de Aperfeicoame...

Abstract. The human mind has been designed to evaluate similarity fast and efficiently. When building/using a data format to make the web content more machine-friendly, can we learn something useful from how the mind represents data? We present four theories psychological theories that tried to solve the problem and how they relate to semantic web practices. Metric models (such as the vector space model and LSA) were the first-comers and still have important advantages. Advances in Bayesian methods pushed Feature models ( e.g., Topic model). Structural mapping models propose that for similarity, shared structure matters more, although the formalisms that express these ideas are still developing. Transformational distance models (e.g., syntagmatic-paradigmatic-SP- model) reduce similarity to information transmission. Topic and SP models do not require preexisting classes but still have a long way to go; the need of automatically generating structure is less pressing when one of the driving forces of the semantic web is the creation of ontologies.