FURI | Spring 2022
Assessing the Effect of Overparameterization in Quantum Neural Networks
Current quantum devices aim to leverage quantum mechanical features in tandem with classical computing resources to iteratively solve high-value problems even the best supercomputers would be incapable of. For instance, one application of quantum neural networks (QNN) is to find the ground state energy of molecules. However, the utility of such algorithms is currently limited, as often the iterative search gets stuck in suboptimal solutions. The importance of overparameterization in classical neural networks motivates the hypothesis that as more parameters are added to the QNN, convergence to globally optimal solutions improves. In this work, overparameterization was numerically analyzed for simple QNNs. It is found that different types of overparameterization can yield similar convergence behavior, which paves the way for characterizing the quantum resources relevant to improving the performance of QNNs. Overall, this research will advance knowledge in the field of quantum machine learning and its applications in chemistry.
Hometown: Henderson, Nevada, United States
Graduation date: Spring 2024