Multilayer feedforward networks are universal approximators
Neural Networks
Neural network design
Hi-index | 0.00 |
People are usually more interested in ranking a group of interrelated objects than the actual scores of them. In this paper, we presented an improved relative evaluation criterion, which is suitable for such task and can solve those problems that are difficult to deal with by common absolute criterion. We put forward the improvement algorithm and realize its function approximation by means of BP algorithm. Furthermore, we tested this criterion with the statistic data of SARS. The experimental results indicated that our improved criterion is effective in comparing a set of correlative objects objectively. Finally, we made a conclusion about the goal and efficiency of relative criterion.