A comprehensive review and future prospects of machine learning algorithms based on contrastive learning
Hongru Ma, Wenying Niu · IET conference proceedings. · 2025
With the rapid development of artificial intelligence technology, machine learning algorithms have demonstrated strong application potential in many fields. However, the high dependence of traditional supervised learning on labeled data limits its practical application scenarios. As an emerging unsupervised learning paradigm, contrastive learning has attracted widespread attention in recent years by mining the inherent similarities and differences in data for feature representation learning. This paper systematically combs the theoretical foundation, core algorithms and application status in computer vision, natural language processing, cross-modal tasks and other fields, deeply analyzes the current challenges and future development directions. By summarizing the technical principles and innovation points of representative algorithms such as SimCLR, MoCo, BYOL, SwAV, and DenseCL, etc.), and combining the latest research progress, this paper aims to provide a comprehensive technical reference and research direction guidance for researchers.