Recognition And Application Of Multiple Classifier Under The Distributed Environment

The classification is basic problem of the machine study and pattern-recognition, but with the changes of technology and realistic environment, the traditional categorised method based on single joint has already received the great challenge, need to improve it in order to meet demands of reality. To the deficiency of the traditional recognition system, we proposed the new algorithm in this paper. The groundwork of this text is to structure a lot of classifier base on Boosting algorithm under the distributed environment, and merge Boosting and distributed to for an organic whole ,this paper deals with some primary explorations based on multiple sites and multiple classifier, such as analytical methods, framework design and so on.Firstly, Standard Boosting algorithm is introduced in detail, and we proposed the improving algorithm. A recognition method of multiple classifiers base on distributed Boosting algorithm has designed and realized. In the system using this method, Feature extracting theory is discussed comprehensively. The dataset of the system is pretreated, and the high dimensions data is projected onto low dimensions data so as to raise the recognition ratio. As the basic classifier algorithm of Boosting, the BP neural network is also discussed in detail, and a rational neural network configuration is made successfully which is suitable for Boosting. In the last, we treat the research about multiple classifiers under the distributed environment, such as design, construction and improving, etc. Then we verified systematic dependability further by testing it with the public data aggregate made very effective. It has reached the requirement designed systematically. At the same time, recognition time is obvious reducing.