Sparse Null space LDA for object recognition


In this paper, we improved the classical Linear Discriminant Analysis (LDA) by choosing a group of sparse orthonormal basis to be the discriminative vectors. Unlike the traditional null space LDA which merely focus on a group of orthonormal basis, we constrain the discriminative vectors to be sparse. In this case, each of the vectors merely contains part of information of the original feature of observations so that the projection of observation on the LDA space can be better classified by finding the most similar discriminative vectors.

3 Figures and Tables

Cite this paper

@article{Tian2017SparseNS, title={Sparse Null space LDA for object recognition}, author={Jinyu Tian and Taiping Zhang}, journal={2017 IEEE 2nd International Conference on Cloud Computing and Big Data Analysis (ICCCBDA)}, year={2017}, pages={316-320} }