Quantum algorithm for robust optimization via stochastic-gradient online learning

Debbie Lim,João F. Doriguello,Patrick Rebentrost

2023-04-05

Abstract:Optimization theory has been widely studied in academia and finds a large variety of applications in industry. The different optimization models in their discrete and/or continuous settings has catered to a rich source of research problems. Robust convex optimization is a branch of optimization theory in which the variables or parameters involved have a certain level of uncertainty. In this work, we consider the online robust optimization meta-algorithm by Ben-Tal et al. and show that for a large range of stochastic subgradients, this algorithm has the same guarantee as the original non-stochastic version. We develop a quantum version of this algorithm and show that an at most quadratic improvement in terms of the dimension can be achieved. The speedup is due to the use of quantum state preparation, quantum norm estimation, and quantum multi-sampling. We apply our quantum meta-algorithm to examples such as robust linear programs and robust semidefinite programs and give applications of these robust optimization problems in finance and engineering.

Quantum Physics,Optimization and Control

What problem does this paper attempt to address?

The paper aims to address the following issues:

1. **Proposing Quantum Algorithms to Accelerate Robust Optimization**: This paper proposes a quantum algorithm for online robust optimization, aiming to enhance the speed of classical algorithms by using quantum technologies such as quantum state preparation, quantum norm estimation, and quantum multi-sampling. For a wide range of stochastic subgradients, this quantum algorithm can achieve up to quadratic speedup.

2. **Improving the Meta-Algorithm Proposed by Ben-Tal et al.**: The authors extend the online robust optimization meta-algorithm proposed by Ben-Tal et al., allowing the use of stochastic subgradients, and demonstrate that this extension can maintain the same guarantees as the original non-stochastic version in certain cases.

3. **Specific Application Examples**: The paper demonstrates how this quantum meta-algorithm can be applied to specific robust linear programming and robust semidefinite programming problems, and explores the practical applications of these robust optimization problems in finance and engineering fields, such as the Global Maximum Return Portfolio (GMRP) problem and the Truss Topology Design (TTD) problem.

In summary, this research aims to improve existing robust optimization algorithms by introducing quantum computing methods, providing more efficient solutions for handling large-scale datasets or high-dimensional problems.