A review on multi-view learning

Multi-view learning is gradually becoming a well-established domain within machine learning that tackles problems involving the availability of multiple views or sources of data. Existing multi-view learning reviews mainly focus on a specific task, classifying methods based on their principles or styles.

To solve the problems, a research team led by Zhiwen YU published their new review on 15 July 2025 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.

The team provided a review of multi-view learning from a novel perspective of machine learning paradigms, systematically categorizing existing multi-view learning methods by considering different supervised scenarios and types of tasks.

In the review, they provide a detailed and clear discussion of multi-view learning from multiple aspects, including the basic theory, technology, method categorizations, applications, future development, and challenges of multi-view learning. Specifically, this survey categorizes existing multi-view learning work into four groups: multi-view classification methods, multi-view semi-supervised classification methods, multi-view clustering methods, and multi-view semi-supervised clustering methods. On the basis of these four categories, multi-view classification and multi-view clustering are further divided into three subcategories: multi-view representation learning, incomplete multi-view learning, and the combination of multi-view learning with other machine learning methods. This categorization is based on existing research hotspots and technologies, and deeply analyzes and discusses existing multi-view learning work from the learning paradigm-level (supervised, semi-supervised and unsupervised), task-level (classification, clustering, etc.), data-level (incomplete view, incomplete labels), and technical-level (representation learning, combination with other technologies).

Moreover, they also provide detailed analyses of all groups and the differences between the same subclass for different tasks. This survey provides a comprehensive overview of various aspects of multi-view learning and presents the applications and challenges of multi-view learning in various fields to help researchers better understand the development direction of multi-view learning and their applicable scenarios.