Shapes of features and a modified measure for linear discriminant analysis
15th International Conference on Pattern Recognition (ICPR-2000), Barcelona, İspanya, 3 - 07 Eylül 2000, ss.410-413, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Basıldığı Şehir: Barcelona
- Basıldığı Ülke: İspanya
- Sayfa Sayıları: ss.410-413
- Anahtar Kelimeler: feature selection, feature extraction, selection by shape similarity, Fisher's measure with clustering
- Marmara Üniversitesi Adresli: Hayır
Özet
In this paper, the problem of selecting most representative features among a feature set is considered. Two new feature selection algorithms are introduced and their performances are compared with some well-known feature selection algorithms. The algorithms are tested with the iris data set, three artificially generated data sets and a data set obtained from steel structures.