Benchmarking optimizers and parametrizations for variational quantum optimization with Rydberg atoms

Hybrid quantum-classical algorithms have been a promising class of algorithms in exploring potential quantum advantages using near-term quantum hardware. Selecting an optimizer and a corresponding parameterization of the model is a central step in the algorithm. Different choices of optimization procedures and the parameter space may result in very different performances. However, there is still a lack of systematic benchmarks of different choices of model parameterizations and optimizers. In this work, we benchmark various popular optimizers with hyperparameters search and different model parametrizations including QAOA and quasi-adiabatic algorithms, using the Rydberg Hamiltonian to solve the maximum independent set problem as an example. This benchmark result can be used as heuristics for selecting optimizers and parameterizations for other problems and experimental platforms. Our benchmarks can also be used as the baseline for developing new variational algorithms and optimization procedures.