1: \begin{abstract}

2: \Glspl{VQA} have attracted a lot of attention from the quantum computing community for the last few years. Their hybrid quantum-classical nature with relatively shallow quantum circuits makes them a promising platform for demonstrating the capabilities of \gls{NISQ} devices. Although the classical machine learning community focuses on gradient-based parameter optimization, finding near-exact gradients for \glspl{VQC} with the parameter-shift rule introduces a large sampling overhead. Therefore, gradient-free optimizers have gained popularity in quantum machine learning circles. Among the most promising candidates is the \gls{SPSA} algorithm, due to its low computational cost and inherent noise resilience. We introduce a novel approach that uses the approximated gradient from \gls{SPSA} in combination with state-of-the-art gradient-based classical optimizers. We demonstrate numerically that this outperforms both standard \gls{SPSA} and the parameter-shift rule in terms of convergence rate and absolute error in simple regression tasks. The improvement of our novel approach over \gls{SPSA} with \gls{SGD} is even amplified when shot- and hardware-noise are taken into account. We also demonstrate that error mitigation does not significantly affect our results.

3: \end{abstract}

4: