Multiview Clustering of Adaptive Sparse Representation Based on Coupled P Systems.

Area of Science:

• Data Mining and Machine Learning
• Computational Intelligence

Background:

• Multiview clustering (MVC) is crucial for data mining but often uses fixed neighbors, limiting effectiveness with diverse data.
• Existing MVC methods rely on time-consuming iterative optimization for clustering results.

Purpose of the Study:

• To propose an efficient and effective multiview clustering algorithm without iteration.
• To address the limitations of fixed neighbor selection and iterative optimization in current MVC techniques.

Main Methods:

• Developed a multiview clustering of adaptive sparse representation based on coupled P system (MVCS-CP).
• Employed a parameter-free natural neighbor search for adaptive neighbor determination in each view.
• Utilized manifold learning and sparse representation to construct view-specific similarity matrices.
• Introduced a soft thresholding operator to create a unified graph for clustering.

Main Results:

• The proposed MVCS-CP algorithm demonstrates superior performance compared to state-of-the-art methods.
• Experimental validation on nine real datasets confirms the effectiveness of the MVCS-CP approach.
• The method successfully preserves internal data geometry through similarity matrix construction.

Conclusions:

• MVCS-CP offers an efficient, non-iterative solution for multiview clustering.
• The adaptive neighbor selection and sparse representation enhance clustering accuracy and robustness.
• This approach provides a significant advancement in multiview data analysis.