Journal
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
Volume 33, Issue 1, Pages 130-144Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TNNLS.2020.3027588
Keywords
Discriminant power; joint L-2,L- p- and L-2,L-s-norm; nongreedy; robust discriminant analysis (RDA); rotational invariance; s-norm distance metrics
Categories
Funding
- Central Public-Interest Scientific Institution Basal Research Fund [CAFYBB2019QD003]
- National Science Foundation of China [62072246]
- Natural Science Foundation of Jiangsu Province [BK20171453]
- National Science Foundations of China [61773210, 61773117]
- Qinglan and Six Talent Peaks Projects of Jiangsu Province
Ask authors/readers for more resources
In this article, a new robust discriminant subspace (RDS) learning method is presented for feature extraction. The method uses a different objective function formulation to ensure the subspace is both robust and discriminative. An efficient nongreedy iterative algorithm is proposed to solve the challenging optimization problem. The experimental results on image classification databases demonstrate the effectiveness of RDS.
Recently, there are many works on discriminant analysis, which promote the robustness of models against outliers by using L-1- or L-2,L-1-norm as the distance metric. However, both of their robustness and discriminant power are limited. In this article, we present a new robust discriminant subspace (RDS) learning method for feature extraction, with an objective function formulated in a different form. To guarantee the subspace to be robust and discriminative, we measure the within-class distances based on L-2,L-s-norm and use L-2,L- p-norm to measure the between-class distances. This also makes our method include rotational invariance. Since the proposed model involves both L-2,(p)-norm maximization and L-2,L-s-norm minimization, it is very challenging to solve. To address this problem, we present an efficient nongreedy iterative algorithm. Besides, motivated by trace ratio criterion, a mechanism of automatically balancing the contributions of different terms in our objective is found. RDS is very flexible, as it can be extended to other existing feature extraction techniques. An in-depth theoretical analysis of the algorithm's convergence is presented in this article. Experiments are conducted on several typical databases for image classification, and the promising results indicate the effectiveness of RDS.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available