Abstract

Quantum computers offer a promising route to tackling problems that areclassically intractable such as in prime-factorization, solving large-scalelinear algebra and simulating complex quantum systems, but requirefault-tolerant quantum hardware. On the other hand, variational quantumalgorithms (VQAs) have the potential to provide a near-term route to quantumutility or advantage, and is usually constructed by using parametrized quantumcircuits (PQCs) in combination with a classical optimizer for training.Although VQAs have been proposed for a multitude of tasks such as ground-stateestimation, combinatorial optimization and unitary compilation, there remainmajor challenges in its trainability and resource costs on quantum hardware.Here we address these challenges by adopting Hardware Efficient and dynamicalLIe algebra Supported Ansatz (HELIA), and propose two training schemes thatcombine an existing g-sim method (that uses the underlying group structure ofthe operators) and the Parameter-Shift Rule (PSR). Our improvement comes fromdistributing the resources required for gradient estimation and training toboth classical and quantum hardware. We numerically test our proposal forground-state estimation using Variational Quantum Eigensolver (VQE) andclassification of quantum phases using quantum neural networks. Our methodsshow better accuracy and success of trials, and also need fewer calls to thequantum hardware on an average than using only PSR (upto 60% reduction), thatruns exclusively on quantum hardware. We also numerically demonstrate thecapability of HELIA in mitigating barren plateaus, paving the way for traininglarge-scale quantum models.