Abstract
It takes more time and is easier to fall into the local minimum value when using the traditional full-supervised learning algorithm to train RBFNN. Therefore, the paper proposes one algorithm to determine the RBFNN’s data center based on the improvement density method. First it uses the improved density method to select RBFNN’s data center, and calculates the expansion constant of each center, then only trains the network weight with the gradient descent method. To compare this method with full-supervised gradient descent method, the time not only has obvious reduction (including to choose data center’s time by density method), but also obtains better classification results when using the data set in UCI to carry on the test to the network.
Highlights
T Radial Basis Function Neural Network (RBFNN) is a forward neural network of good performance
The nonlinear ability of RBFNN is mainly reflected in the radial function of the hidden layer whose property is determined by its data center
We proposed an improved density method—a dynamic clustering algorithm in mathematical statistics to select the data center of RBFNN and trained the weight value with gradient descent method
Summary
T Radial Basis Function Neural Network (RBFNN) is a forward neural network of good performance It has faster learning rate than BP neural network and there is no local minimum problem. The nonlinear ability of RBFNN is mainly reflected in the radial function of the hidden layer whose property is determined by its data center. In terms of supervised methods, it needs us to specify the number of the hidden layer. We proposed an improved density method—a dynamic clustering algorithm in mathematical statistics to select the data center of RBFNN and trained the weight value with gradient descent method. This article is organized according to the following contents: Part 2 introduces the basic theory of RBFNN; Part 3 and Part 4 propose the specific algorithm; Part 5 is the experiments and analysis; conclusions
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.