1. Research on Conditioning and Application for the F2N2
- Author
-
Lei Zhou and Zhong Nian Li
- Subjects
Mathematical optimization ,Series (mathematics) ,Linearity error ,business.industry ,General Engineering ,Back propagation neural network ,Nonlinear system ,TheoryofComputation_ANALYSISOFALGORITHMSANDPROBLEMCOMPLEXITY ,Convergence (routing) ,Feedforward neural network ,Conditioning ,Artificial intelligence ,Layer (object-oriented design) ,business ,Mathematics - Abstract
as everyone knows, conditioning F2N2(Feed Forward Neural Network) in many ways, but ways, but has so far failed to completely solve the "convergence not quickly" and "convergence is not robust"of these two problems, thus affecting the application of F2N2. Aiming at this problem, pro this paper puts forward a new method of processing and application of F2N2, the method of flexible BPNN (Back Propagation Neural Network) algorithm and hierarchical optimization algo-rithm as a whole, each layeris independent rights attached to the conditioning process, carefully con-structed F2N2objective function is non linear series based on the description, not only can the optimization of each layer of the weight problem is simplified to a linear problem, but also can effec-tively controlthe linearity error. This paper provides several examples of application of F2N2automa-tic detection, the results of application show that, this method is superior to other methods.
- Published
- 2014