摘要
Dimensionality reduction is very important in pattern recognition, machine learning, and image recognition. In this paper, we propose a novel linear dimensionality reduction technique using trace ratio criterion, namely Discriminant Neighbourhood Structure Embedding Using Trace Ratio Criterion (TR-DNSE). TR-DNSE preserves the local intrinsic geometric structure, characterizing properties of similarity and diversity within each class, and enforces the separability between different classes by maximizing the sum of the weighted distances between nearby points from different classes. Experiments on four image databases show the effectiveness of the proposed approach.
Dimensionality reduction is very important in pattern recognition, machine learning, and image recognition. In this paper, we propose a novel linear dimensionality reduction technique using trace ratio criterion, namely Discriminant Neighbourhood Structure Embedding Using Trace Ratio Criterion (TR-DNSE). TR-DNSE preserves the local intrinsic geometric structure, characterizing properties of similarity and diversity within each class, and enforces the separability between different classes by maximizing the sum of the weighted distances between nearby points from different classes. Experiments on four image databases show the effectiveness of the proposed approach.
