Machine learning models called convolutional neural networks (CNNs) power technologies such as image recognition and language translation. Its quantum counterpart, known as a quantum convolutional neural network (QCNN), can process information more efficiently by using quantum states instead of classical bits.
Photonic systems are a promising platform for QCNN because photons are fast, stable, and easy to manipulate on a chip. However, photonic circuits typically operate linearly, which limits the flexible behavior required by neural networks.
In the study published in Advanced Photonics, researchers introduced a way to increase the adaptability of photonic circuits without sacrificing compatibility with current technology. Their approach adds a controlled step called adaptive state injection, allowing them to adjust the circuit’s behavior based on measurements taken during processing. This additional control brings photonic QCNN closer to practical application.
The team built a modular QCNN using single photons from a quantum dot source and two integrated quantum photonic processors. Similar to traditional CNNs, the network processes information in stages. After the first stage, a portion of the optical signal is measured. Depending on the result, the system either injects new photons or sends existing light forward, gently manipulating the calculations. Today’s photonic hardware cannot switch light in real time without losing information, so the researchers emulated this step in the lab using a controlled technique that reproduces the same effect.
To test their design, they encoded a simple 4 × 4 image, a pattern of horizontal or vertical bars. Measurements at each stage were consistent with theoretical predictions. In the full experimental setup, QCNN achieved a classification accuracy of over 92 percent, consistent with numerical simulations. This shows the potential of an adaptive approach.
The researchers also investigated scalability, noting that future photonic devices with fast switching could enable larger and more powerful QCNNs than some classical methods.
“This study provides both a theoretical framework and a proof-of-concept implementation of a photonic QCNN,” said senior author Fabio Sciarrino. “We hope that these results will serve as a starting point for the development of new quantum machine learning methods.”
This study outlines a realistic path to higher-performance optical quantum processors by adding simple adaptation steps that work with existing technologies.
For more information, see the original Gold Open Access article “Photonic quantum convolutional neural network with adaptive state injection” by L. Monbroussou et al., Adv. Photon. 7(6), 066012 (2025), doi 10.1117/1.AP.7.6.066012
/Open to the public. This material from the original organization/author may be of a contemporary nature and has been edited for clarity, style, and length. Mirage.News does not take any institutional position or position, and all views, positions, and conclusions expressed herein are those of the authors alone. Read the full text here.
