Design And Implementation Of A Distributed Neural Computing Platform

Artificial neuron nerwork(ANN) is builded to mimic the structure and function of animal neuron group. Because of its ability of parallel computation, anti-erroring, generalization, it provides a very effective way to solve the unknown non-liner problems. And now ANN is being wisely used in many different scientific domain, it has much means to develop a service platform providing high performance neurocomputation.This platform applies the concept of distributed computation. Computation platform is required for being used on different platforms and the ANN algorithms must run after simple Employment. Based on this point , we find that the J2EE platform provided fits very well. J2EE is an architecture which is used to simplify the complications of development, deployment and management of enterprise solutions. We present a distributed application system based on J2EE and it makes the whole system become a multi-tier component system, loose-couple horizontally and vertically, provide flexibility, reusability, testability and extensibility.The paper consists of five chapters:Chapter 1 introduced the background and principle of ANN and distributed computation.Chapter 2 compared several software architectures, and then introduced J2EE application platform.Chapter 3 described the design and realization of the platform.Chapter 4 discussed the optimization of BP modal using genetic algorithm.In Chapter 5, we summarize the whole thesis.