همایش بین المللی هوش مصنوعی

صفحه اصلی / دومین همایش بین المللی هوش مصنوعی

Photonic Quantum Hardware for AI Optimization Tasks

نویسندگان :

Sarah Daneshzad¹ Gholam-Mohammad Parsanasab²

1- Shahid Beheshti University, Tehran, Iran 2- Shahid Beheshti University, Tehran, Iran

کلمات کلیدی :

photonic quantum computing،Quantum Approximate Optimization Algorithm (QAOA)،integrated photonics،AI optimization،quantum hardware platforms

چکیده :

Recent advances in photonic quantum computing offer a promising route for accelerating combinatorial optimization problems central to artificial intelligence (AI). This paper presents a hardware-aware simulation framework for implementing the Quantum Approximate Optimization Algorithm (QAOA) on photonic hardware platforms, extending prior photonic QAOA implementations for the Max-Cut problem to more general AI-relevant graph optimization tasks. Unweighted, weighted, and balanced graph-partitioning problems are modeled as representative benchmarks, and their cost and mixer unitaries are compiled into integrated-photonics primitives based on Mach–Zehnder interferometers, programmable phase shifters, and cross-Kerr-type interactions. Depth-\mathbit{p} photonic QAOA circuits \left(\mathbit{p}=\mathbf{1}\mathrm{-}\mathbf{3}\right) are simulated using the Strawberry Fields photonic backend with realistic models of propagation loss, phase noise, and detector inefficiency. The results show monotonic improvement in approximation ratio with increasing depth, achieving up to \mathbf{\rho}\approx\mathbf{0}.\mathbf{87} for graphs with \mathbit{n}\ \le\mathbf{18} and less than 10% degradation under nominal noise conditions. Warm-start initialization reduces the number of optimization iterations by approximately 30% compared with random initialization. Benchmark comparisons with Goemans–Williamson and simulated annealing demonstrate that photonic QAOA attains near-classical performance for small- to medium-sized graphs, indicating that integrated photonics is a feasible platform for near-term quantum optimization.