Abstract:Non-negative matrix factorization (NMF) approximates a non-negative matrix $X$ by a product of two non-negative low-rank factor matrices $W$ and $H$. NMF and its extensions minimize either the Kullback-Leibler divergence or the Euclidean distance between $X$ and $W^T H$ to model the Poisson noise or the Gaussian noise. In practice, when the noise distribution is heavy tailed, they cannot perform well. This paper presents Manhattan NMF (MahNMF) which minimizes the Manhattan distance between $X$ and $W^T H$ for modeling the heavy tailed Laplacian noise. Similar to sparse and low-rank matrix decompositions, MahNMF robustly estimates the low-rank part and the sparse part of a non-negative matrix and thus performs effectively when data are contaminated by outliers. We extend MahNMF for various practical applications by developing box-constrained MahNMF, manifold regularized MahNMF, group sparse MahNMF, elastic net inducing MahNMF, and symmetric MahNMF. The major contribution of this paper lies in two fast optimization algorithms for MahNMF and its extensions: the rank-one residual iteration (RRI) method and Nesterov's smoothing method. In particular, by approximating the residual matrix by the outer product of one row of W and one row of $H$ in MahNMF, we develop an RRI method to iteratively update each variable of $W$ and $H$ in a closed form solution. Although RRI is efficient for small scale MahNMF and some of its extensions, it is neither scalable to large scale matrices nor flexible enough to optimize all MahNMF extensions. Since the objective functions of MahNMF and its extensions are neither convex nor smooth, we apply Nesterov's smoothing method to recursively optimize one factor matrix with another matrix fixed. By setting the smoothing parameter inversely proportional to the iteration number, we improve the approximation accuracy iteratively for both MahNMF and its extensions.

Chordal-NMF with Riemannian Multiplicative Update

Relative Pairwise Relationship Constrained Non-Negative Matrix Factorisation

NeNMF: an Optimal Gradient Method for Nonnegative Matrix Factorization.

Adaptive graph regularized nonnegative matrix factorization for data representation

Spatially correlated nonnegative matrix factorization

Robust Structured Nonnegative Matrix Factorization for Image Representation.

MahNMF: Manhattan Non-negative Matrix Factorization

Convergence Analysis of Graph Regularized Non-Negative Matrix Factorization

Non-Negative Matrix Factorization with Sinkhorn Distance.

Cauchy balanced nonnegative matrix factorization

Factor-Bounded Nonnegative Matrix Factorization

Sum-of-norms regularized Nonnegative Matrix Factorization

Large-Cone Nonnegative Matrix Factorization

Nonnegative Matrix Factorization on Orthogonal Subspace

Correntropy Supervised Non-Negative Matrix Factorization

Adaptive computation of the Symmetric Nonnegative Matrix Factorization (NMF)

Efficient algorithms for regularized Poisson Non-negative Matrix Factorization

Spatially correlated nonnegative matrix factorization for image analysis

Non-negative Matrix Factorization on Manifold

Optimization and expansion of non-negative matrix factorization

Stratified-NMF for Heterogeneous Data