ENHANCING THE EFFECTIVE UTILIZATION OF NOISY QUANTUM COMPUTERS THROUGH STREAMLINING ARCHITECTURE AND ALGORITHMIC IMPROVEMENTS

Abstract

Quantum computing emerges as a powerful tool for complex calculations, poised to reshape computing by performing tasks beyond the reach of classical systems. Unfortunately, quantum computers exhibit inherent noise from various sources. This dissertation aims to enhance the effective utilization of noisy quantum computers through advances in three layers of the quantum computation stack. The author optimized a quantum neural network for network anomaly detection on current quantum computers, achieving an F1 score of 0.86, surpassing comparable studies. A novel metric, the Certainty Factor, is introduced to analyze noise susceptibility in quantum classifiers and enrich predictions with uncertainty measures. Additionally, the author accelerated parametrized quantum circuit instantiation through the design and implementation of QFactor-Sample, a domain-specific optimization algorithm enabling a 2,000x speedup over popular optimizers. This improvement enhances the tradeoff between runtime and result quality. Lastly, a new mathematical framework was developed for analyzing quantum circuits and error models without relying on Monte Carlo techniques. The author utilized it to provide a detailed error model for various implementations of critical circuits in fault-tolerant quantum computation, demonstrating the framework’s generality, efficiency, and potential contribution to optimizing quantum computer architectures.