Scalable On-Hardware Training of Quantum Neural Networks and Application to Clinical Data Imputation
Researchers have developed a new framework for training quantum neural networks (QNNs) on quantum hardware, significantly reducing the computational cost of gradient estimation. This method lowers the required circuit evaluations from quadratic to logarithmic in the number of qubits, making QNN optimization practical for larger systems. The framework was successfully applied to clinical data imputation using the MIMIC-III dataset, with models trained on IonQ hardware demonstrating performance comparable to or exceeding classical baselines while showing reduced variance. AI
IMPACT Enables more practical and scalable training of quantum neural networks for real-world applications.