A regularization method for quantum neural networks using data symmetry

Abstract

Leveraging data symmetries has recently become a key strategy in quantum neural networks (QNNs) to improve generalization and training efficiency. In this study, we propose a novel regularization method for QNNs based on input data symmetry. By introducing a penalty term that encourages the model to align with data symmetry, our method enables improved training speed and generalization. This symmetry-based regularization is simple to implement and does not require prior knowledge of the symmetry group. We validate its effectiveness through numerical experiments on both classification tasks and quantum generative adversarial networks. Empirical results demonstrate faster convergence and lower test errors. Furthermore, we provide a theoretical generalization bound using Rademacher complexity and conjecture a condition under which models with symmetry exhibit better generalization. Our findings highlight the potential of symmetry-aware regularization in enhancing the performance of QML models.