Clustering via Self-Supervised Diffusion

2025-07-06

AI Summary

This work introduces diffusion models to unsupervised clustering for the first time, addressing the longstanding challenge of jointly optimizing feature discriminability and clustering robustness. We propose DiffCluster: a framework that extracts high-dimensional semantic features using a pretrained Vision Transformer (ViT), then employs a self-supervised diffusion model as a teacher network to generate diverse, structure-preserving pseudo-labels via iterative denoising. A student network distills these outputs to achieve stable cluster assignments. Crucially, we reinterpret the diffusion process as an implicit data augmentation mechanism and an uncertainty-aware clustering prior, significantly enhancing resilience to input noise, intra-class variation, and complex manifold structures. Evaluated on standard benchmarks—including CIFAR-10/100 and ImageNet-Dogs—DiffCluster achieves state-of-the-art performance, improving average clustering accuracy by 3.2% while demonstrating superior generalization and robustness.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Clustering
• Search and Optimization: Learning to Search

Application Category

• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Diffusion models, widely recognized for their success in generative tasks, have not yet been applied to clustering. We introduce Clustering via Diffusion (CLUDI), a self-supervised framework that combines the generative power of diffusion models with pre-trained Vision Transformer features to achieve robust and accurate clustering. CLUDI is trained via a teacher-student paradigm: the teacher uses stochastic diffusion-based sampling to produce diverse cluster assignments, which the student refines into stable predictions. This stochasticity acts as a novel data augmentation strategy, enabling CLUDI to uncover intricate structures in high-dimensional data. Extensive evaluations on challenging datasets demonstrate that CLUDI achieves state-of-the-art performance in unsupervised classification, setting new benchmarks in clustering robustness and adaptability to complex data distributions.

Problem

• Applying diffusion models to clustering tasks for the first time
• Combining diffusion models with Vision Transformer for robust clustering
• Enhancing clustering accuracy via stochastic teacher-student data augmentation

Innovation

• Self-supervised diffusion model for clustering
• Teacher-student paradigm with stochastic diffusion
• Combines diffusion models with Vision Transformer