Abstract:We consider the problem of simultaneously clustering and learning a linear representation of data lying close to a union of low-dimensional manifolds, a fundamental task in machine learning and computer vision. When the manifolds are assumed to be linear subspaces, this reduces to the classical problem of subspace clustering, which has been studied extensively over the past two decades. Unfortunately, many real-world datasets such as natural images can not be well approximated by linear subspaces. On the other hand, numerous works have attempted to learn an appropriate transformation of the data, such that data is mapped from a union of general non-linear manifolds to a union of linear subspaces (with points from the same manifold being mapped to the same subspace). However, many existing works have limitations such as assuming knowledge of the membership of samples to clusters, requiring high sampling density, or being shown theoretically to learn trivial representations. In this paper, we propose to optimize the Maximal Coding Rate Reduction metric with respect to both the data representation and a novel doubly stochastic cluster membership, inspired by state-of-the-art subspace clustering results. We give a parameterization of such a representation and membership, allowing efficient mini-batching and one-shot initialization. Experiments on CIFAR-10, -20, -100, and TinyImageNet-200 datasets show that the proposed method is much more accurate and scalable than state-of-the-art deep clustering methods, and further learns a latent linear representation of the data.

A Class of Manifold Regularized Multiplicative Update Algorithms for Image Clustering

Convergence Analysis of Graph Regularized Non-Negative Matrix Factorization

Efficient Algorithms of Box-Constrained Nonnegative Matrix Factorization and Its Applications in Image Clustering

Learning Manifold Structures With Subspace Segmentations

Adaptive graph regularized nonnegative matrix factorization for data representation

Sketch-based multiplicative updating algorithms for symmetric nonnegative tensor factorizations with applications to face image clustering

Class-driven nonnegative matrix factorization with manifold regularization for data clustering

Image Clustering Based On Graph Regularized Robust Principal Component Analysis

Penalized Nonnegative Nonnegative Matrix Tri-Factorization For Co-Clustering

Single multiplicatively updated matrix factorization for co-clustering.

Adaptive Manifold Regularized Matrix Factorization for Data Clustering.

A Novel Data Representation Framework Based on Nonnegative Manifold Regularisation

Centric graph regularized log-norm sparse non-negative matrix factorization for multi-view clustering

Semi-supervised sparse subspace clustering with manifold regularization

Correntropy-Based Low-Rank Matrix Factorization With Constraint Graph Learning for Image Clustering

A Regularized Approach for Geodesic-Based Semisupervised Multimanifold Learning

Multi-view clustering based on pairwise co-regularization and robust dual graph non-negative matrix factorization

Unsupervised Manifold Linearizing and Clustering

Robust Manifold Matrix Factorization for Joint Clustering and Feature Extraction.

Robust non-negative matrix factorization with multiple correntropy-induced hypergraph regularizer

Graph Regularized Sparse Non-Negative Matrix Factorization for Clustering