Multi-view L2-SVM and its multi-view core vector machine.

Area of Science:

• Machine Learning
• Computer Science

Background:

• Multi-view classification presents challenges in integrating information from diverse data sources.
• Existing methods may lack flexibility or scalability for complex datasets.

Purpose of the Study:

• To introduce a novel L2-SVM based classifier, Multi-view L2-SVM, for multi-view classification tasks.
• To develop an efficient version, MvCVM, for large-scale datasets based on Generalized Core Vector Machine (GCVM).

Main Methods:

• Proposed Multi-view L2-SVM classifier with an unbiased objective function for controlled support vector generation.
• Imposed consensus among multiple views to exploit coherence and differences.
• Extended Multi-view L2-SVM to MvCVM using GCVM for efficient large-scale training.

Main Results:

• Demonstrated the effectiveness of Multi-view L2-SVM on small-scale multi-view datasets.
• Showcased the efficiency of MvCVM for large-scale multi-view datasets, featuring asymptotic linear time and constant space complexity.
• Experimental results validated the proposed classifiers' performance.

Conclusions:

• The proposed Multi-view L2-SVM classifier effectively addresses multi-view classification challenges.
• MvCVM provides a scalable and efficient solution for large-scale multi-view data analysis.
• Both methods demonstrate significant improvements in classification performance.