Subspace Clustering on Incomplete Data with Self-Supervised Contrastive Learning
This addresses the problem of subspace clustering in real-world scenarios with missing data for applications like computer vision and recommendation systems, representing an incremental improvement over existing methods.
The paper tackles subspace clustering on incomplete data by proposing Contrastive Subspace Clustering (CSC), which uses self-supervised contrastive learning to generate masked views and learn invariant embeddings, then clusters them with sparse subspace clustering; experiments on six benchmarks show it outperforms classical and deep learning baselines with strong robustness to missing data and scalability.
Subspace clustering aims to group data points that lie in a union of low-dimensional subspaces and finds wide application in computer vision, hyperspectral imaging, and recommendation systems. However, most existing methods assume fully observed data, limiting their effectiveness in real-world scenarios with missing entries. In this paper, we propose a contrastive self-supervised framework, Contrastive Subspace Clustering (CSC), designed for clustering incomplete data. CSC generates masked views of partially observed inputs and trains a deep neural network using a SimCLR-style contrastive loss to learn invariant embeddings. These embeddings are then clustered using sparse subspace clustering. Experiments on six benchmark datasets show that CSC consistently outperforms both classical and deep learning baselines, demonstrating strong robustness to missing data and scalability to large datasets.