Frame of pattern recognition based on multi-Agent
Xiaobo Chen · Caai Transactions on Intelligent Systems · 2006
Traditional pattern recognition methods have two disadvantages in general:centralized control and depending on the quality of the features extracted or selected.APRF(Agent-Based Pattern Recognition Frame)is proposed based on multi-agent theories and techniques after the analysis of the characteristics,advantages and disadvantages of the top-down and bottom-up methods.The pattern is modeled by quantitatively computing from bottom to top so as to ease the classification.Then the pattern is emerged by the qualitatively analytical approach from top to bottom.APRF is aimed to make the pattern recognition by computer more accordant with the cognition of human.The core of APRF is:Pattern Recognition=Pattern Modeling+Pattern Emergence.