An on-chip photonic deep neural network for image classification

• Published in: Nature (London) Vol. 606; no. 7914; pp. 501 - 506
• Main Authors: Ashtiani, Farshid, Geers, Alexander J., Aflatouni, Firooz
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 16.06.2022; Nature Publishing Group
• Subjects: 639/166/987; 639/624; 639/624/399/1099; Access time; Analog to digital conversion; Artificial neural networks; Classification; Computation; Computer vision; Deep learning; Energy efficiency; Handwriting; Humanities and Social Sciences; Image classification; Integrated circuits; Light; Machine learning; multidisciplinary; Neural networks; Neurons; Optical communication; Optics; Photonics; Science; Science (multidisciplinary)
• ISSN: 0028-0836; 1476-4687
• DOI: 10.1038/s41586-022-04714-0

Deep neural networks with applications from computer vision to medical diagnosis 1 – 5 are commonly implemented using clock-based processors 6 – 14 , in which computation speed is mainly limited by the clock frequency and the memory access time. In the optical domain, despite advances in photonic computation 15 – 17 , the lack of scalable on-chip optical non-linearity and the loss of photonic devices limit the scalability of optical deep networks. Here we report an integrated end-to-end photonic deep neural network (PDNN) that performs sub-nanosecond image classification through direct processing of the optical waves impinging on the on-chip pixel array as they propagate through layers of neurons. In each neuron, linear computation is performed optically and the non-linear activation function is realized opto-electronically, allowing a classification time of under 570 ps, which is comparable with a single clock cycle of state-of-the-art digital platforms. A uniformly distributed supply light provides the same per-neuron optical output range, allowing scalability to large-scale PDNNs. Two-class and four-class classification of handwritten letters with accuracies higher than 93.8% and 89.8%, respectively, is demonstrated. Direct, clock-less processing of optical data eliminates analogue-to-digital conversion and the requirement for a large memory module, allowing faster and more energy efficient neural networks for the next generations of deep learning systems. Using a three-layer opto-electronic neural network, direct, clock-less sub-nanosecond image classification on a silicon photonics chip is demonstrated, achieving a classification time comparable with a single clock cycle of state-of-the-art digital implementations.