Discriminative clustering via extreme learning machine.

Area of Science:

• Machine Learning
• Data Mining
• Pattern Recognition

Background:

• Discriminative clustering integrates supervised classification principles into unsupervised learning.
• Effective clustering aims for data partitions that are easily classifiable.
• Existing methods may not fully leverage discriminative information for clustering.

Purpose of the Study:

• To propose novel discriminative clustering algorithms.
• To enhance clustering performance by maximizing data discrimination.
• To leverage Extreme Learning Machine (ELM) and Fisher's Linear Discriminant Analysis (LDA).

Main Methods:

• Developed three discriminative clustering algorithms based on ELM.
• Algorithm 1: Iterative training of weighted ELM (W-ELM) to maximize discrimination.
• Algorithms 2 & 3: Based on Fisher's LDA, using alternative optimization and kernel k-means.

Main Results:

• Proposed algorithms are easy to implement.
• Achieved competitive clustering accuracy on real-world datasets.
• Demonstrated superior performance compared to state-of-the-art clustering methods.

Conclusions:

• Discriminative clustering based on ELM and LDA offers a powerful approach.
• The proposed methods effectively improve clustering by enhancing data discrimination.
• These algorithms provide a practical and effective solution for real-world clustering tasks.