Domain Adaptation Based on Semi-Supervised Cross-Domain Mean Discriminative Analysis and Kernel Transfer Extreme...

Area of Science:

• Computer Science
• Machine Learning
• Pattern Recognition

Background:

• Traditional feature extraction and classification models degrade when training and testing data distributions mismatch.
• Domain shift is a significant challenge in real-world pattern recognition tasks.

Purpose of the Study:

• To propose a novel domain adaptation approach to address the performance degradation caused by domain shift.
• To enhance the robustness and accuracy of pattern recognition across different data distributions.

Main Methods:

• Introduced cross-domain mean approximation (CDMA) into semi-supervised discriminative analysis (SDA) to develop semi-supervised cross-domain mean discriminative analysis (SCDMDA) for shared feature extraction.
• Utilized kernel extreme learning machine (KELM) as a classifier and developed a kernel transfer extreme learning machine (KTELM) by incorporating a cross-domain mean constraint for improved knowledge transfer.

Main Results:

• The proposed SCDMDA and KTELM methods demonstrated superior performance compared to existing state-of-the-art techniques.
• Experiments on four real-world cross-domain visual datasets validated the effectiveness of the approach.

Conclusions:

• The developed domain adaptation strategy effectively handles discrepancies in data distributions.
• The proposed method offers a competitive and robust solution for cross-domain pattern recognition challenges.