A PATTERN CLASSIFIER - MODIFIED AFC, AND HANDWRITTEN DIGIT RECOGNITION

We modified the adaptive fuzzy classification algorithm (AFC), which a llows fuzzy clusters to grow to meet the demands of a given task durin g training. Every fuzzy cluster is defined by a reference vector and a fuzzy cluster radius, and it is represented as a shape of hypersphere in pattern space. Any pattern class is identified by overlapping plur al hyperspherical fuzzy clusters so that it is possible to approximate complex decision boundaries among pattern classes. The modified AFC w as applied to recognize handwritten digits, and performances were show n compared with other neural networks.