CONCEPT LEARNING AND FORM RECOGNITION

点击翻译按钮获取中文摘要

The long-range goal of this research is to develop a coherent quantitative theory of (visual) pattern recognition that will deal in detail with the role of learning in this process. To avoid complexities such as the possibility that visual scenes or patterns may be recognized on the basis of remembered verbal descriptions (among other reasons), the experiments will be done with pigeons as subjects. The ability to recognize patterns is prerequisite to any significant interaction with either the physical or social environment. The proposed research is organized about a quantitative model that already accounts for a large number of phenomena observed in experiments on acquisition of choice behavior involving stimuli that vary along one or more prothetic continua. The experiments use relatively simple visual patterns, such as letters or geometric forms, displayed on the face of a TV monitor. They are concerned with the way organisms learn to discriminate among such patterns, how the patterns are represented in memory, with the decision processes involved in the recognition of these patterns, and with the role of generalization in the abstraction of form. All the work is done by methods that require the pigeons to classify the visual patterns presented to them into two or more categories.