Quantum computing holds great potential for advancing the limitations of machine learning algorithms to handle higher data dimensions and reduce overall training parameters in deep neural network (DNN) models. This study uses a parameterized quantum circuit (PQC) on a gate-based quantum computer to investigate the potential for quantum advantage in a model-free reinforcement learning problem. Through a comprehensive investigation and evaluation of the current model and capabilities of quantum computers, we designed and trained a novel hybrid Quantum neural network based on the latest Qiskit and PyTorch framework. We compared its performance with a full-classical DNN with and without an integrated PQC. Our research provides insights into the potential of deep quantum learning to solve a maze problem and, potentially, other reinforcement learning problems. We conclude that various reinforcement learning problems can be effective with reasonable training epochs. Moreover, a comparative discussion of the various quantum reinforcement learning model on maze problems is discussed to evaluate our research's overall potential and advantages.

翻译：摘要：量子计算有望突破机器学习算法的局限性，以处理更高维度的数据并减少深度神经网络（DNN）模型中的总训练参数。本研究在基于门控量子计算机上使用参数化量子电路（PQC），探索无模型强化学习问题中量子优势的可能性。通过对当前模型及量子计算机能力的全面调查与评估，我们基于最新的Qiskit和PyTorch框架设计并训练了一种新型混合量子神经网络。我们将其性能与未集成PQC的全经典DNN及集成PQC的DNN进行了对比。研究揭示了深度量子学习在解决迷宫问题（以及潜在的其它强化学习问题）方面的潜力。我们得出结论：在合理的训练轮次下，各类强化学习问题均可有效求解。此外，本文通过对比讨论迷宫问题中多种量子强化学习模型的表现，评估了本研究的整体潜力与优势。