Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a 15-minute conference presentation introducing AccelerQ, a novel framework that automatically optimizes quantum eigensolver implementations using machine learning and search-based optimization techniques. Learn how this approach treats quantum eigensolver programs as black-box systems and employs XGBoost regression models combined with genetic algorithms to efficiently tune hyperparameters for improved accuracy and computational efficiency. Discover the framework's two key innovations: training on data from smaller, classically simulable quantum systems and using program-specific machine learning models that exploit persistent local physical interactions in molecular systems to enable generalization to larger quantum systems. Examine empirical evaluation results on two distinct quantum eigensolver implementations - ADAPT-QSCI and QCELS - where models trained on systems up to 16 qubits successfully optimized hyperparameters for larger 20-, 24-, and 28-qubit Hamiltonians, achieving error reductions from 5.48% to 5.05% for ADAPT-QSCI and from 7.5% to 6.5% for QCELS. Understand how this research demonstrates the potential for integrating software engineering methodologies with quantum computing applications, offering promising directions for quantum software stack optimization when direct classical simulation becomes computationally impractical.
