Quantum phases classification using quantum machine learning with SHAP-driven feature selection

Abstract We introduce a novel methodology for the classification of quantum phases within the ANNNI (axial next-nearest-neighbor Ising) model by employing quantum machine learning (QML) techniques in conjunction with the Shapley additive explanations (SHAP) algorithm for effective feature selection. The study is centered on two eminent quantum algorithms: quantum support vector machines (QSVM) and variational quantum classifiers (VQC). Our findings demonstrate that both QSVM and VQC attain notable predictive accuracy when confined to merely 5 or 6 critical features, indicating that an optimized subset of observables can maintain robust performance with reduced computational cost. In examining the relationship between accuracy and the number of features, we demonstrated that the incorporation of supplementary features may even negatively impact the model’s accuracy. The proposed framework underscores the potential of interdisciplinary strategies for the analysis of intricate quantum systems, furthering progress in the application of quantum machine learning to recognize patterns in quantum information science.