A STRUCTURE-PARAMETER-ADAPTIVE (SPA) NEURAL TREE FOR THE RECOGNITION OF LARGE CHARACTER SET

An adaptive self-organizing neural tree called the structure-parameter-adaptive (SPA) neural tree is presented in this paper. The SPA neural tree can adapt to a changing environment parametrically and structurally. The SPA neural tree grows when new concepts are learned and shrinks when old concepts are disposed. It also makes parameter adaptation to cope with small changes in the environment. The SPA neural tree is suited to hierarchical classification. Its application to character pattern recognition is also presented in this paper. When applied to character pattern recognition, it shows promising performance.