Anti-lopsided Algorithm for Large-scale Nonnegative Least Square Problems

Non-negative least squares (NNLS) problem is one of the most important fundamental problems in numeric analysis. It has been widely used in scientific computation and data modeling. In big data, the limitations of algorithm speed and accuracy are typical challenges. In this paper, we propose fast and robust anti-lopsided algorithm with high accuracy that is totally based on the first order methods. The main idea of our algorithm is to transform the original NNLS into an equivalent non-negative quadratic programming, which significantly reduce the scaling problem of variables. The proposed algorithm can reach high accuracy and fast speed with linear convergence $(1 - \frac{1}{2||Q||_2})^k$ where $\sqrt{n} \leq ||Q||_2 \leq n$, and $n$ is the dimension size of solutions. The experiments on large matrices clearly show the high performance of the proposed algorithm in comparison to the state-of-the-art algorithms.

No reviews yet.

No comments yet.