سال انتشار: ۱۳۸۴

محل انتشار: یازدهمین کنفرانس سالانه انجمن کامپیوتر ایران

تعداد صفحات: ۶

نویسنده(ها):

A. Mirzaei – Dept. of Computer Engineering Amirkabir UniversityValiasr,Tehran, Iran, 15914
M. Rahmati – Dept. of Computer Engineering Amirkabir UniversityValiasr,Tehran, Iran, 15914
R. Safabahsh –

چکیده:

A new method for extracting discriminative features is represented in this paper. Most of the ordinary feature extractionalgorithms are limited to the use of linear transformation approach. However, in most practical classification problems it is required to focus our attention on nonlinear functions. Consequently, the goal of this research is to introduce a new feature extraction method which can extract nonlinear functions without any apriori knowledge about the form of the function. In order to achieve this, a neural network structure is introduced which the inputs are the original features and the outputs are the enhanced features. Experimental results indicate that our proposed method is able to extract features which are superior to conventional methods.