Human concept learning presents a version of the classic problem of induction, which is made particularly difficult by the combination of two requirements: the need to learn from a rich (i.e. nested and overlapping) vocabulary of possible concepts and the need to be able to generalize concepts reasonably from only a few positive examples. I begin this thesis by considering a simple number concept game as a concrete illustration of this ability. On this task, human learners can with reasonable confidence lock in on one out of a billion billion billion logically possible concepts, after seeing only four positive examples of the concept, and can generalize informatively after seeing just a single example. Neither of the two classic approaches to inductive inference -- hypothesis testing in a constrained space of possible rules and computing similarity to the observed examples -- can provide a complete picture of how people generalize concepts in even this simple setting. This thesis prop...

Various diagrams have been used heavily in systems analysis and design without proper verification of their usability. However, different diagrammatic representations of the same information may vary in the computational efficiency of working with these diagrams. The objective of this research was to explore the effects of diagrammatic representations on the task of integrating multiple diagrams. The domain of systems analysis and design was used to generate examples and test the theory. A cognitive model of diagram integration was proposed, and an experimental study was conducted to explore the effects of representational features of diagrams on the cognitive process of diagram integration. Results of the experiment show that the representational features of the diagrams acted as the criteria for selecting among various methods for analyzing and designing the integrated diagram. In addition, the difference in the selected methods resulted in different task performances in terms of ana...

Abstract not found

One of the difficulties of object recognition stems from the need to overcome the variability in object appearance caused by factors such as illumination and pose. The influence of these factors can be countered by learning to interpolate between stored views of the target object, taken under representative combinations of viewing conditions. Difficulties of another kind arise in daily life situations that require categorization, rather than recognition, of objects. We show that, although categorization cannot rely on interpolation between stored examples, knowledge of several representative members, or prototypes, of each of the categories of interest can still provide the necessary computational substrate for the categorization of new instances. The resulting representational scheme based on similarities to prototypes is computationally viable, and is readily mapped onto the mechanisms of biological vision revealed by recent psychophysical and physiological studies. 1 Introduction T...

Abstract not found

this technical term was introduced by Helmholtz (in a German equivalent), and despite variations in theorists' characterizations of such inferences, which are noted as needed. To give some sense of the range of theories, I begin by briefly examining two areas: size and distance perception, and color constancy

The paper compares predictions derived from the similaritybased and the theory-based accounts of young children's induction. The former predicts the primacy of induction from one single entity to another single entity (one-to-one induction), whereas the latter does not predict such primacy. Predictions were tested in three experiments where 4-5 yearolds and 11-12 year-olds were asked to perform inductive generalization of biological properties. Participants could generalize properties either from a single animal to another single animal (one-to-one induction) or from a group of animals to a single animal (many-to-one induction). Experiments 1 and 2 revealed that under various stimuli presentation conditions, young children exhibited a strong preference of one-to-one induction, performing generalizations in a similarity-based manner. At the same time, preadolescents exhibited a strong preference of manyto -one induction, performing generalizations in a theorybased manner. In Experiment 3, an alternative explanation that one-to-one induction stems from a tendency to match quantifiers or label endings was tested and eliminated. Results are discussed in relation to cognitive and developmental aspects of inductive inference.

We monitored regional cerebral activity with BOLD fMRI during acquisition of a novel semantic category and subsequent categorization of test stimuli by a rule-based strategy or a similarity-based strategy. We observed different patterns of activation in direct comparisons of rule- and similarity-based categorization. During rule-based category acquisition, subjects recruited anterior cingulate, thalamic, and parietal regions to support selective attention to perceptual features, and left inferior frontal cortex to helps maintain rules in working memory. Subsequent rule-based categorization revealed anterior cingulate and parietal activation while judging stimuli whose conformity with the rules was readily apparent, and left inferior frontal recruitment during judgments of stimuli whose conformity was less apparent. By comparison, similarity-based category acquisition showed recruitment of anterior prefrontal and posterior cingulate regions, presumably to support successful retrieval of previously encountered exemplars from long-term memory, and bilateral temporal-parietal activation for perceptual feature integration. Subsequent similaritybased categorization revealed temporal–parietal, posterior cingulate, and anterior prefrontal activation. These findings suggest that largescale networks support relatively distinct categorization processes

The aim of this paper is to present a software artifact for machine supported understanding and modelling of prototypebased categorization. This software system is able to perform discovery of syndromes (seen as prototypes) into a given data set of clinical observations. A new genetic algorithm is used to achieve an unsupervised categorization of observation via adaptive learning of number and features of prototypes. Its evolutionary learning is oriented to maximize specificity and distinctiveness of categories. Experimental results show that prototype-based categorization of clinical observation is suitable for syndrome-based categorization. The trichotomy of categorizations (superordinate, basic and subordinate) is explained by trade off between specificity and distinctiveness. Moreover the natural basic level is also related to epistemic adequacy of found prototypes.

Abstract: The structure of words is often thought to provide important evidence regarding the structure of concepts. At the same time, most contemporary linguists posit a great deal of structure in words. Such a trend makes some atomists about concepts uncomfortable. The details of linguistic methodology undermine several strategies for avoiding positing structure in words. I conclude by arguing that there is insufficient evidence to hold that word-structure bears any interesting relation to the structure of concepts.