Correntropy-Based Hypergraph Regularized NMF for Clustering and Feature Selection on Multi-Cancer Integrated Data.

Area of Science:

• Machine Learning
• Bioinformatics
• Data Science

Background:

• Non-negative Matrix Factorization (NMF) is a key technique for clustering and feature selection.
• Traditional NMF struggles with noisy data, outliers, and unaddressed data manifold structures.
• Existing methods lack robustness and fail to fully utilize high-order data relationships.

Purpose of the Study:

• To introduce a novel correntropy-based hypergraph regularized NMF (CHNMF) method.
• To enhance the robustness and accuracy of NMF in the presence of noise and outliers.
• To leverage high-order geometric information within datasets for improved clustering and feature selection.

Main Methods:

• Developed CHNMF by incorporating correntropy into the loss term for increased robustness.
• Integrated a hypergraph regularization term to capture complex, high-order data structures.
• Employed the half-quadratic (HQ) optimization technique to efficiently solve the CHNMF objective function.

Main Results:

• The proposed CHNMF method demonstrated superior performance compared to existing state-of-the-art techniques.
• Experiments on multi-cancer integrated data validated the effectiveness of CHNMF.
• CHNMF showed significant improvements in both data clustering and feature selection tasks.

Conclusions:

• CHNMF offers a robust and effective alternative to traditional NMF for data analysis.
• The method's ability to handle noise and capture intricate geometric patterns is a key advantage.
• CHNMF shows promise for applications in complex biological and other data-intensive fields.