Regensburg 2022 – wissenschaftliches Programm
QI 4.28: Poster
Montag, 5. September 2022, 18:00–20:00, P2
Performances and limitations of variational quantum algorithms under realistic noise models — •Marco Schumann, Frank Wilhelm-Mauch, and Alessandro Ciani — Institute for Quantum Computing Analytics (PGI-12), Forschungszentrum Jülich, 52425 Jülich, Germany
As the field of quantum computing progresses, an intriguing question is if the currently available NISQ devices can deliver a quantum advantage for practical applications, like Variational Quantum Algorithms (VQAs). Due to the lack of error correction, noise limits the performance of these algorithms. Recently it was shown that a quantum state undergoing a variational quantum circuit with general Pauli noise approaches the completely mixed state with increasing circuit depth. In this ongoing project, we study the question of how variational quantum circuits behave under more realistic noise models, like dephasing or amplitude damping noise. We use the Quantum Approximate Optimization Algorithm (QAOA), which is a specific VQA, to solve combinatorial optimization problems. Considering the problem of MaxCut on different d-regular graphs with different qubit numbers, we run the circuit for many instances of randomly chosen circuit parameters. We find that for weak amplitude damping or dephasing noise the average purity of the output state of the circuit approaches the purity of the completely mixed state while the variance of the purity approaches zero with increasing circuit depth. The decrease of the average purity is well described by an exponential decay, where the decay rate is approximately linear in the noise strength.