Regularized correntropy criterion based semi-supervised ELM.

Area of Science:

• Machine Learning
• Data Mining
• Artificial Intelligence

Background:

• Semi-supervised learning (SSL) is crucial for leveraging large unlabeled datasets.
• Extreme Learning Machines (ELMs) offer efficient and accurate data classification.
• Standard ELM optimization using Mean Square Error (MSE) struggles with non-Gaussian noise.

Purpose of the Study:

• To develop a robust Semi-Supervised Extreme Learning Machine (SSELM) resistant to non-Gaussian noise.
• To enhance the accuracy and learning efficiency of SSELM through a novel optimization criterion.
• To introduce a new method, the robust regularized correntropy criterion based SSELM (RC-SSELM).

Main Methods:

• Developed a robust regularized correntropy criterion based SSELM (RC-SSELM).
• Optimized the output weight matrix using a fixed-point iteration approach.
• Analyzed the convergence of RC-SSELM via half-quadratic optimization techniques.

Main Results:

• RC-SSELM demonstrated superior performance compared to standard SSELM and other state-of-the-art methods.
• Experimental results on synthetic and UCI benchmark datasets validated the effectiveness of RC-SSELM.
• The proposed method showed improved robustness in the presence of non-Gaussian noise.

Conclusions:

• RC-SSELM offers a more robust and accurate approach to semi-supervised learning.
• The correntropy criterion effectively addresses limitations of MSE in noisy environments.
• This work advances the field of efficient and accurate data classification using SSL.