Hybrid Quantum-Classical Photonic Neural Networks

TL;DR

This paper introduces hybrid quantum-classical photonic neural networks that leverage quantum speedup to enhance computational capacity and accuracy without increasing physical network size, promising advancements in AI hardware.

Contribution

It demonstrates that combining classical layers with quantum circuits improves neural network performance and scalability in integrated photonic systems.

Findings

Hybrid networks match larger classical networks in accuracy.

Hybrid networks maintain performance with reduced bit precision.

Quantum integration enhances network scalability.

Abstract

Neuromorphic (brain-inspired) photonics leverages photonic chips to accelerate artificial intelligence, offering high-speed and energy efficient solutions in RF communication, tensor processing, and data classification. However, the limited physical size of integrated photonic hardware constrains network complexity and computational capacity. In light of recent advances in photonic quantum technology, it is natural to utilize quantum exponential speedup to scale photonic neural network capabilities. Here we show a combination of classical network layers with trainable continuous variable quantum circuits yields hybrid networks with improved trainability and accuracy. On a classification task, hybrid networks achieve the same performance when benchmarked against fully classical networks that are twice the size. When the bit precision of the optimized networks is reduced through added noise, the hybrid networks still achieve greater accuracy when evaluated at state of the art bit precision. These hybrid quantum classical networks demonstrate a unique route to improve computational capacity of integrated photonic neural networks without increasing the physical network size.