English

A Faster Interior Point Method for Semidefinite Programming

Data Structures and Algorithms 2020-09-23 v1 Optimization and Control

Abstract

Semidefinite programs (SDPs) are a fundamental class of optimization problems with important recent applications in approximation algorithms, quantum complexity, robust learning, algorithmic rounding, and adversarial deep learning. This paper presents a faster interior point method to solve generic SDPs with variable size n×nn \times n and mm constraints in time \begin{align*} \widetilde{O}(\sqrt{n}( mn^2 + m^\omega + n^\omega) \log(1 / \epsilon) ), \end{align*} where ω\omega is the exponent of matrix multiplication and ϵ\epsilon is the relative accuracy. In the predominant case of mnm \geq n, our runtime outperforms that of the previous fastest SDP solver, which is based on the cutting plane method of Jiang, Lee, Song, and Wong [JLSW20]. Our algorithm's runtime can be naturally interpreted as follows: O~(nlog(1/ϵ))\widetilde{O}(\sqrt{n} \log (1/\epsilon)) is the number of iterations needed for our interior point method, mn2mn^2 is the input size, and mω+nωm^\omega + n^\omega is the time to invert the Hessian and slack matrix in each iteration. These constitute natural barriers to further improving the runtime of interior point methods for solving generic SDPs.

Keywords

Cite

@article{arxiv.2009.10217,
  title  = {A Faster Interior Point Method for Semidefinite Programming},
  author = {Haotian Jiang and Tarun Kathuria and Yin Tat Lee and Swati Padmanabhan and Zhao Song},
  journal= {arXiv preprint arXiv:2009.10217},
  year   = {2020}
}

Comments

FOCS 2020