Research on Service-oriented NN Autonomous Ensemble

To resolve the running time and computing resource problems of neural networks processing large-scale data set, based on the analysis of Neural Networks' distributed learning algorithm and Autonomous Decentralized System theory, a Neural Networks Autonomous Ensemble model and its integration mechanism are proposed in this paper. The autonomous integration model is based on Autonomous Decentralized System, its autonomous coordination is achieved by mobile agents, push mobile agents accomplish the data allocation, and pull mobile agents collect the networks' result and evaluate them, all these are realized via Faded Information Field, the system's on-line ability and unit autonomy is assured, and the experimental results show that system's generalization and classification ability is assured.