Parameter-Free Multiview K-Means Clustering With Coordinate Descent Method.

Area of Science:

• Data Science
• Machine Learning
• Clustering Algorithms

Background:

• Real-world datasets often contain heterogeneous features from multiple sources.
• Existing multiview clustering methods struggle with hyper-parameter tuning and high computational demands.
• There is a need for improved clustering performance in multiview data analysis.

Purpose of the Study:

• To develop a novel, parameter-free multiview clustering framework (PFMKM).
• To address the limitations of existing methods, including hyper-parameter sensitivity and computational complexity.
• To enhance the accuracy and efficiency of clustering heterogeneous data.

Main Methods:

• Introduced a parameter-free multiview -means clustering with coordinate descent (PFMKM).
• Employed a self-weighted scheme to learn feature weights, eliminating manual hyper-parameter tuning.
• Developed an efficient coordinate descent optimization algorithm for direct cluster indicator matrix calculation.

Main Results:

• PFMKM demonstrated superior performance compared to state-of-the-art multiview clustering methods.
• The parameter-free nature significantly simplifies the application of the method.
• The coordinate descent optimization reduced computational complexity and improved clustering accuracy.

Conclusions:

• PFMKM offers an effective and efficient solution for multiview clustering on heterogeneous data.
• The parameter-free design makes it a practical choice for real-world applications.
• The method achieves competitive or superior results, validating its approach.