Abstract:Clustering algorithm is one of the most widely used and influential analysis techniques. With the advent of deep learning, deep embedding clustering algorithms have rapidly evolved and yield promising results. Much of the success of these algorithms depends on the potential expression captured by the autoencoder network. Therefore, the quality of the potential expression directly determines the algorithm's performance. In view of this, researchers have proposed many improvements. Although the performance has been slightly improved, they all have one shortcoming, that is, too much emphasis is placed on the original data reconstruction ability during the process of feature expression, which greatly limits the further expression of potential features according to specific clustering tasks. Moreover, there is a large amount of noise in the original data, so blindly emphasizing reconstruction will only backfire. Hence, we innovatively propose a deep embedding clustering algorithm based on residual autoencoder (DECRA) after in-depth research. Specifically, a novel autoencoder network with residual structure is proposed and introduced into deep embedded clustering tasks. The network introduces an adaptive weight layer in feature representation z, which can make it have good robustness, generalization for specific tasks, and adaptive learning of better feature embeddings according to category classification. In this paper, the reasons for the validity of this structure are explained theoretically, and comprehensive experiments on six benchmark datasets including various types show that the clustering performance of the DECRA is very competitive and significantly superior to the most advanced methods.

Contrastive Deep Embedded Clustering

Semi-supervised deep embedded clustering.

Deep Robust Clustering by Contrastive Learning

Unsupervised discriminative feature learning via finding a clustering-friendly embedding space

Doubly Contrastive Deep Clustering

Deep Embedding Clustering Based on Residual Autoencoder

Contrastive Deep Convolutional Transform K-Means Clustering

Deep Structural Contrastive Subspace Clustering.

Deep subspace clustering to achieve jointly latent feature extraction and discriminative learning

Improving Deep Embedded Clustering via Learning Cluster-level Representations.

Deep clustering based on embedded auto-encoder

Deep Temporal Contrastive Clustering

Unsupervised Deep Embedding for Clustering Analysis

Deep cross-modal subspace clustering with Contrastive Neighbour Embedding

Contrastive graph clustering via enhanced hard sample mining and cluster-guiding

Effective Sample Pairs Based Contrastive Learning for Clustering

Deep Clustering by Semantic Contrastive Learning

Deep Discriminative Clustering Analysis

Deep Contrastive Graph Learning with Clustering-Oriented Guidance

Deep Embedded Clustering with Asymmetric Residual Autoencoder

Unsupervised Contrastive Clustering Via Density Cluster Representative Combination