A recognition and association systems based on primitive concept formation using three layered neural networks are proposed. This system forms prototype and abstract concepts from experiences that resulted in similar actions on a simple structure. By expressing an image as an array of ones and zeros representing the presence or absence of micro-features, the reactions between images can be simply defined as modified Boolian operations on the features of each image. "Multiplication," an operation that extracts the common features of two images, generates abstract concepts from concrete images.
展开▼