Quantum Circuit Synthesis via Reinforcement Learning: Automated Design of Efficient Quantum Algorithms

Keywords: quantum computing, reinforcement learning, circuit synthesis, quantum algorithms, automated optimization, quantum supremacy, gate decomposition, hardware-aware design

Abstract: Quantum circuit synthesis remains a critical bottleneck in quantum computing, requiring expert knowledge to translate high-level algorithms into hardware-efficient implementations. This paper introduces QLSynth, a reinforcement learning framework that automates quantum circuit design by treating gate sequences as policy actions and reward functions as fidelity/performance metrics. Our approach achieves 30-50% reduction in gate count and 40% lower circuit depth compared to state-of-the-art synthesis tools while maintaining >99% fidelity. We demonstrate efficacy across quantum Fourier transform, Grover's search, and Shor's factoring algorithms, showing adaptability to different qubit topologies and noise profiles. This work represents a paradigm shift in quantum software development, moving from manual design to AI-driven automated optimization.

Submission Number: 232