Online Nonnegative Matrix Factorization with Outliers
This work addresses large-scale data processing with outliers, but it is incremental as it builds on existing nonnegative matrix factorization methods.
The authors tackled the problem of online nonnegative matrix factorization with outliers by proposing a unified framework with two solvers, proving convergence and demonstrating computational efficiency and efficacy in tasks like basis learning and image denoising.
We propose a unified and systematic framework for performing online nonnegative matrix factorization in the presence of outliers. Our framework is particularly suited to large-scale data. We propose two solvers based on projected gradient descent and the alternating direction method of multipliers. We prove that the sequence of objective values converges almost surely by appealing to the quasi-martingale convergence theorem. We also show the sequence of learned dictionaries converges to the set of stationary points of the expected loss function almost surely. In addition, we extend our basic problem formulation to various settings with different constraints and regularizers. We also adapt the solvers and analyses to each setting. We perform extensive experiments on both synthetic and real datasets. These experiments demonstrate the computational efficiency and efficacy of our algorithms on tasks such as (parts-based) basis learning, image denoising, shadow removal and foreground-background separation.