Bootstrap Deep Spectral Clustering with Optimal Transport

2025/08/06 by Guo, Wengang, Ye, Wei, Chen, Chunchun +4 · 1 citation
#Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2508.04200

Abstract

Spectral clustering is a leading clustering method. Two of its major shortcomings are the disjoint optimization process and the limited representation capacity. To address these issues, we propose a deep spectral clustering model (named BootSC), which jointly learns all stages of spectral clustering -- affinity matrix construction, spectral embedding, and k-means clustering -- using a single network in an end-to-end manner. BootSC leverages effective and efficient optimal-transport-derived supervision to bootstrap the affinity matrix and the cluster assignment matrix. Moreover, a semantically-consistent orthogonal re-parameterization technique is introduced to orthogonalize spectral embeddings, significantly enhancing the discrimination capability. Experimental results indicate that BootSC achieves state-of-the-art clustering performance. For example, it accomplishes a notable 16% NMI improvement over the runner-up method on the challenging ImageNet-Dogs dataset. Our code is available at https://github.com/spdj2271/BootSC.

Citations

• Deep Spectral Clustering via Joint Spectral Embedding and Kmeans
• Image Clustering Conditioned on Text Criteria
• Large Language Models Enable Few-Shot Clustering
• ClusterLLM: Large Language Models as a Guide for Text Clustering
• DivClust: Controlling Diversity in Deep Clustering
• Investigating and Mitigating the Side Effects of Noisy Views for Self-Supervised Clustering Algorithms in Practical Multi-View Scenarios
• GPT-4 Technical Report
• Air Quality in the New Delhi Metropolis under COVID-19 Lockdown
• HCSC: Hierarchical Contrastive Selective Coding
• Deep Embedded K-Means Clustering
• You Never Cluster Alone
• Clustering-friendly Representation Learning via Instance Discrimination and Feature Decorrelation
• Graph Contrastive Clustering
• OTA: Optimal Transport Assignment for Object Detection
• Learning Transferable Visual Models From Natural Language Supervision
• Exploring Simple Siamese Representation Learning
• Exploring Simple Siamese Representation Learning
• Unsupervised Learning of Visual Features by Contrasting Cluster Assignments
• Online Deep Clustering for Unsupervised Representation Learning
• Acoustic Scene Clustering Using Joint Optimization of Deep Embedding Learning and Clustering Iteration
• SCAN: Learning to Classify Images without Labels
• A Simple Framework for Contrastive Learning of Visual Representations
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Momentum Contrast for Unsupervised Visual Representation Learning
• Momentum Contrast for Unsupervised Visual Representation Learning
• Self-labelling via simultaneous clustering and representation learning
• Deep Spectral Clustering using Dual Autoencoder Network
• Deep Comprehensive Correlation Mining for Image Clustering
• Spectral Clustering via Ensemble Deep Autoencoder Learning (SC-EDAE)
• Invariant Information Clustering for Unsupervised Image Classification and Segmentation
• Deep Clustering for Unsupervised Learning of Visual Features
• Unified Spectral Clustering with Optimal Graph
• Deep Subspace Clustering Networks
• CNN-Based Joint Clustering and Representation Learning with Feature Drift Compensation for Large-Scale Image Data
• Deep Clustering via Joint Convolutional Autoencoder Embedding and Relative Entropy Minimization
• Towards K-means-friendly Spaces: Simultaneous Deep Learning and Clustering
• SGDR: Stochastic Gradient Descent with Warm Restarts
• Mini-Batch Spectral Clustering
• Joint Unsupervised Learning of Deep Representations and Image Clusters
• Deep Residual Learning for Image Recognition
• Unsupervised Deep Embedding for Clustering Analysis
• Optimal Transport for Domain Adaptation
• Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
• A tutorial on spectral clustering
• Least squares quantization in PCM
• A Generalized Solution of the Orthogonal Procrustes Problem
• Contrastive Clustering

Cited by

• On the Provable Importance of Gradients for Language-Assisted Image Clustering

Related

• Deep Spectral Clustering via Joint Spectral Embedding and Kmeans
• Deep Spectral Clustering using Dual Autoencoder Network
• Snapshot Spectral Clustering -- a costless approach to deep clustering ensembles generation
• Spectral Clustering via Ensemble Deep Autoencoder Learning (SC-EDAE)
• Unified Spectral Clustering with Optimal Graph