Deep neural network-based deep clustering has achieved remarkable success by unifying representation learning and clustering. However, conventional representation modules are typically not tailored for clustering, resulting in conflicting objectives that hinder the model’s ability to capture semantic structures with high intra-cluster cohesion and low inter-cluster separation. To overcome this limitation, we propose a novel framework called Triple-stream Contrastive Deep Embedding Clustering via Semantic Structure (TCSS). TCSS is composed of representation and clustering modules, with its innovation rooted in several key designs that ensure their synergistic interaction for modeling semantic structures. First, TCSS introduces a triple-stream input framework that processes the raw instance along with its limited and aggressive augmented views. This design enables a new triple-stream self-training clustering loss, which uncovers implicit cluster structures by contrasting the three input streams. Second, within this loss, a dynamic clustering structure factor is developed to represent the evolving semantic structure in the representation space, thereby constraining the clustering-prediction distribution. Third, TCSS integrates semantic structure-aware techniques, including a clustering-oriented negative sampling strategy and a triple-stream alignment scheme based on k-nearest neighbors and centroids, to refine semantic structures both locally and globally. Extensive experiments on five benchmark datasets demonstrate that TCSS outperforms state-of-the-art methods.

Deep clustering aims to discover meaningful data groups by jointly learning representations and cluster probability distributions. Yet existing methods rarely consider the underlying information characteristics of these distributions, causing ambiguity and redundancy in cluster assignments, particularly when different augmented views are used. To address this issue, this paper proposes a novel information-principled deep clustering framework for learning invariant, redundancy-reduced, and discriminative cluster probability distributions, termed ICIRD. Specifically, ICIRD is built upon three complementary modules for cluster probability distributions: (i) conditional entropy minimization, which increases assignment certainty and discriminability; (ii) inter-cluster mutual information minimization, which reduces redundancy among cluster distributions and sharpens separability; and (iii) cross-view mutual information maximization, which enforces semantic consistency across augmented views. Additionally, a contrastive representation mechanism is incorporated to provide stable and reliable feature inputs for the cluster probability distributions. Together, these components enable ICIRD to jointly optimize both representations and cluster probability distributions in an information-regularized manner. Extensive experiments on five image benchmark datasets demonstrate that ICIRD outperforms most existing deep clustering methods, particularly on fine-grained datasets such as CIFAR-100 and ImageNet-Dogs.