A Survey on Deep Semi-supervised Learning

TL;DR

This survey comprehensively reviews deep semi-supervised learning methods, categorizing existing approaches, comparing their architectures and losses, and discussing current challenges with potential solutions.

Contribution

It provides a detailed taxonomy, reviews 52 methods, and analyzes their differences, offering insights into recent advances and open problems in the field.

Findings

Categorized deep semi-supervised learning methods into five groups.

Compared 52 representative methods in terms of losses and architectures.

Discussed shortcomings and proposed heuristic solutions for open problems.

Abstract

Deep semi-supervised learning is a fast-growing field with a range of practical applications. This paper provides a comprehensive survey on both fundamentals and recent advances in deep semi-supervised learning methods from perspectives of model design and unsupervised loss functions. We first present a taxonomy for deep semi-supervised learning that categorizes existing methods, including deep generative methods, consistency regularization methods, graph-based methods, pseudo-labeling methods, and hybrid methods. Then we provide a comprehensive review of 52 representative methods and offer a detailed comparison of these methods in terms of the type of losses, contributions, and architecture differences. In addition to the progress in the past few years, we further discuss some shortcomings of existing methods and provide some tentative heuristic solutions for solving these open problems.