Quantum Circuit Training with Growth-Based Architectures
This addresses the challenge of balancing expressivity and stability in quantum scientific machine learning applications, representing an incremental improvement over existing methods.
The study tackled the problem of overfitting and model complexity in parameterized quantum circuits by introducing growth-based training strategies that incrementally increase circuit depth during training. The result showed that these dynamic growth methods outperformed traditional fixed-depth approaches, achieving lower final losses and reduced variance between runs in regression tasks and the 2D Laplace equation.
This study introduces growth-based training strategies that incrementally increase parameterized quantum circuit (PQC) depth during training, mitigating overfitting and managing model complexity dynamically. We develop three distinct methods: Block Growth, Sequential Feature Map Growth, and Interleave Feature Map Growth, which add reuploader blocks to PQCs adaptively, expanding the accessible frequency spectrum of the model in response to training needs. This approach enables PQCs to achieve more stable convergence and generalization, even in noisy settings. We evaluate our methods on regression tasks and the 2D Laplace equation, demonstrating that dynamic growth methods outperform traditional, fixed-depth approaches, achieving lower final losses and reduced variance between runs. These findings underscore the potential of growth-based PQCs for quantum scientific machine learning (QSciML) applications, where balancing expressivity and stability is essential.