### Current browse context:

math.OC

### Change to browse by:

### References & Citations

# Mathematics > Optimization and Control

# Title: Boosting Method in Approximately Solving Linear Programming with Fast Online Algorithm

(Submitted on 8 Jul 2021)

Abstract: In this paper, we develop a new algorithm combining the idea of ``boosting'' with the first-order algorithm to approximately solve a class of (Integer) Linear programs(LPs) arisen in general resource allocation problems. Not only can this algorithm solve LPs directly, but also can be applied to accelerate the Column Generation method. As a direct solver, our algorithm achieves a provable $O(\sqrt{n/K})$ optimality gap, where $n$ is the number of variables and $K$ is the number of data duplication bearing the same intuition as the boosting algorithm. We use numerical experiments to demonstrate the effectiveness of our algorithm and several variants.

Link back to: arXiv, form interface, contact.