Different Aspects for Enhancing The Backpropagation Neural Networks | ||||
| Journal of the ACS Advances in Computer Science | ||||
| Article 6, Volume 1, Issue 1, 2007, Page 63-79 PDF (1.41 MB) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/asc.2007.147562 | ||||
 View on SCiNiTO | ||||
| Abstract | ||||
| Backpropagation (BP) algorithm is one of the most popular training algorithms for multilayer neural networks. The convergence of backpropagation learning is analyzed so as to explain common phenomenon observed by specialists. The performance of the backpropagation algorithm is studied, analysed and evaluated in this paper. A method for accelerating the convergence rate is presented. It provides useful guidelines for thinking about how to accelerate the convergence through learning rate adaptation. This work has been implemented through computer simulated using C# with different activation functions and different methods for representing the learning rates. The obtained results are encourage and promising. | ||||
| Keywords | ||||
| Artificial Neural Network; Backpropagation; Activation Functions; Learning Rates; Momentum | ||||
|
Statistics Article View: 81 PDF Download: 148 |
||||
