A Comprehensive Survey on Deep Clustering: Taxonomy, Challenges, and Future Directions

By Sheng Zhou et al

Published on June 15, 2022

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Preliminary
3. Representation Learning Module
3.1 Auto-Encoder based Representation Learning
3.2 Deep Generative Representation Learning
3.3 Mutual Information Maximization Representation Learning
3.4 Contrastive Representation Learning
3.5 Clustering Friendly Representation Learning
3.6 Subspace Representation Learning
3.7 Data type specific representation learning
4. Clustering Module
5. Interactions between Representation Learning and Clustering
6. Benchmark Datasets and Evaluation Metrics
7. Applications of Deep Clustering
8. Limitations, Challenges, and Future Research Directions

Summary

Clustering is a fundamental machine learning task that has been widely studied in the literature. Classic clustering methods follow the assumption that data are represented as features in a vectorized form through various representation learning techniques. As data become increasingly complicated and complex, traditional clustering methods cannot handle the high-dimensional data type. With the success of deep learning, especially deep unsupervised learning, many representation learning techniques with deep architectures have been proposed. The concept of Deep Clustering, i.e., jointly optimizing the representation learning and clustering, has attracted growing attention in the community. Motivated by the success of deep learning in clustering, a comprehensive survey on deep clustering is conducted in this paper, proposing a new taxonomy of different state-of-the-art approaches. The survey categorizes existing methods by the ways they design interactions between deep representation learning and clustering. It also provides popular benchmark datasets, evaluation metrics, and open-source implementations to illustrate various experimental settings. Practical applications of deep clustering and challenging topics deserving further investigations as future directions are discussed.