Overview

---

Quantum Approximate Optimization Algorithm (QAOA), is a hybrid quantum-classical algorithm that can use quantum computers to find approximate solutions for certain combinatorial optimization problems such as Max-Cut or 3-SAT. It is implemented using a number of parameterized unitary operators with an angle value for each operator (2p total parameters, p>1).

In Noisy Intermediate-Scale Quantum (NISQ) era, QAOA algorithm—like many other quantum algorithms—is subject to different types of noise when applied in today's quantum computers.

In this research one of my interns and I used a Reinforcement Learning approach to improve the performance of QAOA in presence of coherent and incoherent quantum noises.

We first implemented an OpenAI-based environment that simulated running a QAOA algorithm for a Max-Cut problem in presence of several coherent and incoherent quantum noises. We then used an Actor-Critic RL algorithm to solve the problem and compared the performance with a baseline implementation running in the same environment.