Robust Subspace Clustering With Complex Noise.

Area of Science:

• Computer Vision
• Pattern Recognition
• Machine Learning

Background:

• Subspace clustering is vital for analyzing high-dimensional data in computer vision and pattern recognition.
• Existing methods struggle with complex noise, which has non-Gaussian and non-Laplacian statistical structures.
• Large corruptions from complex noise are not well-handled by current iterative sparse representation techniques.

Purpose of the Study:

• To propose a novel robust subspace clustering optimization model.
• To address the limitations of existing methods in handling complex noise and large corruptions.
• To improve the accuracy and robustness of subspace clustering.

Main Methods:

• A new optimization model with a two-part objective function: sparse representation and maximizing correntropy.
• Correntropy is utilized as a robust measure to mitigate the impact of large corruptions.
• An extension of pairwise link constraints is incorporated as prior information.
• Half-quadratic minimization is employed for efficient solution of the proposed formulations.

Main Results:

• The proposed method demonstrates superior performance compared to state-of-the-art subspace clustering algorithms.
• Experimental results on three benchmark datasets validate the effectiveness of the correntropy-based approach.
• The method successfully suppresses the influence of large corruptions caused by complex noise.

Conclusions:

• The novel robust subspace clustering model effectively handles complex noise and large corruptions.
• Maximizing correntropy offers a significant advantage in robustly learning low-dimensional subspace structures.
• The proposed method represents a significant advancement in robust subspace clustering for computer vision and pattern recognition applications.