量子地球科学是一门崭新的跨学科前缘专业,量子计算和量子机器学习算法为地学大数据的深度挖掘与分析带来了新的契机。其中,量子神经网络是目前最具代表性的研究方向之一,在复杂多源数据处理方面的效率与准确率尤为突出。本文以大地构造环境判别这一关键问题为切入点,利用堆叠集成算法对量子神经网络(Stacking Quantum Neural Network, S-QNN)进行了改进,并分别实现了玄武岩、辉长岩和尖晶石的构造环境智能判别;同时与四种传统算法(SVM、RF、KNN和NB)、经典神经网络(ANN)和传统量子神经网络(QNN)进行对比。结果表明,集成后的S-QNN模型在3类情况下的准确率较最优的传统算法分别提升5.67%、6.19%和13.34%,较普通的QNN模型提升3.11%、4.99%和3.84%,且更具鲁棒性和通用性。该研究反映了所提出的S-QNN在数据处理中的优势,更证实了量子机器学习算法在地球科学研究中的适用性与潜力,为量子科学与地球科学的交叉融合提供了新思路。

Quantum geoscience represents a cutting-edge interdisciplinary field that leverages quantum computing and quantum machine learning algorithms to revolutionize the analysis of geological data. Among these advancements, the quantum neural network stands out for its efficiency and accuracy in processing complex multi-source data. This study focuses on addressing the challenge of discriminating tectonic settings, enhancing the quantum neural network (S-QNN) with an ensemble strategy to differentiate between basalt, gabbro, and spinel settings. Comparative analyses are conducted with four traditional algorithms (SVM, RF, KNN, NB), artificial neural network (ANN), and traditional quantum neural network (QNN). Results demonstrate that the S-QNN model outperforms the optimal traditional algorithm by 5.67%, 6.19%, and 13.34% in the respective cases, and surpasses the QNN by 3.11%, 4.99%, and 3.84%. The S-QNN model exhibits robustness and versatility, highlighting its superiority in data processing. This study underscores the potential of quantum machine learning algorithms in geoscience research, showcasing the advantages of S-QNN and paving the way for innovative integration of quantum science and geoscience.