A study on the effect of scaling functions to feature weighting performance

In this paper, we perform a study on several data scaling functions for feature weighting. In our former study, we have proposed a feature weighting method based on the Localized Generalization Error Model (L-GEM). The function of weighting those inputs is influential to the performance of resulting classifiers. However, there are few researches focusing on how to use feature weights in a better way. In this paper, we study data scaling function for automatic image annotation with Radial Basis Function Neural Network (RBFNN). Experimental results show that a good data scaling functions yields a better image annotation performance for the same set of feature weights.