Label-Consistent Dictionary Learning via Generalized Adaptive Weighted Recursive Least Squares Algorithm

Articles in Press

Document Type : Original Article

Authors

1 Engineering Department, University of Guilan, Rasht, Iran

2 University of Guilan

10.22034/jasp.2025.65206.1265

Abstract

The generalized adaptive weighted recursive least squares (GAW-RLS) algorithm has demonstrated exceptional performance in data representation within unsupervised dictionary learning methods. This paper extends the GAW-RLS algorithm to supervised settings for dictionary updating. In the proposed method, the cost function incorporates a novel condition termed discriminant sparse representation error, which is added to the representation error and classification error. The dictionary, discriminant sparse representation parameter, and classifier parameter are jointly trained to enhance image classification performance. This approach introduces the Label-Consistent Generalized Adaptive Weighted Recursive Least Squares (LC-GAWRLS) algorithm, which leverages an additional correction weight to regulate the influence of new training data during the dictionary updating process. This enhancement improves robustness against variations in training data compared to existing dictionary learning methods. Simulation results confirm that the LC-GAWRLS algorithm achieves superior classification accuracy, particularly in scenarios with limited training data, demonstrating its potential for effective supervised learning in challenging conditions.

Keywords

Main Subjects

Advanced Signal Processing

Articles in Press, Accepted Manuscript
Available Online from 06 September 2025

Files

Share

How to cite

Statistics