Optimal Design, Based On Bp Neural Network Estimation Of Distribution Algorithms

Nowadays BP Neural Network is the most widely applied Neural Network, the mirroring capacity ensures the clear-cut classification, either from complex or simple respective. However, traditional BP Algorithm has some defects, e.g., slow speed of convergence rate induces local minimum. All these give a limit to the application of BP Neural Network.Estimation of Distribution Algorithm is an emerging algorithm in Graduate Field of Evolution, which has a strong global searching capacity and robustness. it can be applied in optimization of Neural Network, it will definitely bring in quite positive results.The paper introduces the defects of the BP Algorithm, and concludes with some method improvements. Then introduce some estimation of distribution algorithm and propose some method improvements in EDA. Then it applies the Estimation of Distribution Algorithm in optimization of Neural Network, concentrating on the design scheme of algorithm. The paper compares BP Algorithm with Estimation of Distribution Algorithm in mainly two respective: convergence speed and optimization function. The comparison indicates that estimation of distribution algorithm get the better of BP Algorithm in optimization.