Subspace clustering based on locality regularized latent low-rank representation with Frobenius norm minimization

Subspace clustering has garnered significant attention owing to its marvelous effectiveness across diverse tasks, including pattern recognition and computer vision. Among the current subspace clustering methods, latent low-rank representation stands as a highly potent technique. The Frobenius norm based subspace clustering algorithm utilizing latent low-rank representations (FLLRR) is an improvement of the LLRR algorithm when the process of singular value decomposition is particularly time-intensive. However, FLLRR only considers the global structure and does not take into account the local geometric information and the situation where the data may be affected by multiple noises. In this paper, we present a locality regularized FLLRR model (LR-FLLRR) tailored for subspace clustering challenges. Then, we devise an accelerated alternating direction multiplier method (AADMM) to efficiently solve the model. Finally, we perform experiments using several real datasets. The experimental outcomes demonstrate that our method has higher clustering accuracy than several advanced approaches.