Artificial Intelligence in a Photonic Temporal Processor
Abstract
Optical neural networks (ONNs) promise high-throughput and energy-efficient artificial intelligence, yet essentially all implementations so far encode information across space either in free-space arrays or in integrated waveguide meshes, tying the number of neurons to the number of physical components and fixes the routing topology at fabrication. Here we show that moving the computation into time decouples computational dimension from hardware dimension. Exploiting space-time duality, we implement optical diffraction and interference entirely in time domain, using thin-film lithium niobate modulators as time lenses and temporal masks, with chromatic dispersion providing the coupling between successive temporal neurons. We experimentally verify high-order, complex-valued matrix-matrix multiplications using just a single optical input/output port, scaling the computational dimensions far beyond the channel count. By incorporating optical feedback, we extend this platform into versatile neural networks, where the network layers, neuron numbers, and synaptic connections are fully programmable and in-situ trainable. Our temporal diffractive neural networks are successfully validated on various classification benchmarks, alongside image and video generation tasks. Notably, using this platform we demonstrate an all-analogue generative pipeline in which the latent variable is drawn directly from amplified spontaneous emission, so that no digital sampling or electronic modulation appears anywhere in the generative path. Furthermore, high-resolution images and videos are generated at high frame rates, outperforming state-of-the-art modulator-refresh-limited optical generative systems. These results establish a unified photonic temporal computing framework, providing a scalable and deployable pathway toward next-generation machine intelligence.
Related
- Transfer learning-aided, artificial intelligence approach to evaluate optical properties of materials for classical and quantum photonic applications
- Programmable Non-Hermitian Synchronization of Light on a Silicon Photonic Processor
- Universal Broadband Linear Optical Transformations by an Interlaced Structured Integrated Photonic Processor
- Programmable nonlinear function synthesis on a photonic processor with quantum signal processing
- In-orbit operation of a programmable quantum photonic processor
BibTeX
@misc{chen2026artificial,
title = {{Artificial Intelligence in a Photonic Temporal Processor}},
author = {Youlve Chen and Jinlong Xiang and Yimin Hu and Yuchen Yin and Chaojun Xu and Zhengshun Lei and Xin Wang and Yufeng Zhang and Yixiao Zhu and Qunbi Zhuge and Junwen Zhang and Wei Chu and Tao Lin and Yikai Su and Zhipei Sun and Xuhan Guo},
howpublished = {arXiv},
year = {2026},
month = aug,
doi = {10.48550/arxiv.2608.29469},
eprint = {2608.29469},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.29469},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Artificial Intelligence in a Photonic Temporal Processor
AU - Youlve Chen
AU - Jinlong Xiang
AU - Yimin Hu
AU - Yuchen Yin
AU - Chaojun Xu
AU - Zhengshun Lei
AU - Xin Wang
AU - Yufeng Zhang
AU - Yixiao Zhu
AU - Qunbi Zhuge
AU - Junwen Zhang
AU - Wei Chu
AU - Tao Lin
AU - Yikai Su
AU - Zhipei Sun
AU - Xuhan Guo
PY - 2026
DA - 2026/08/29
PB - arXiv
DO - 10.48550/arxiv.2608.29469
UR - https://arxiv.org/abs/2608.29469
AB - Optical neural networks (ONNs) promise high-throughput and energy-efficient artificial intelligence, yet essentially all implementations so far encode information across space either in free-space arrays or in integrated waveguide meshes, tying the number of neurons to the number of physical components and fixes the routing topology at fabrication. Here we show that moving the computation into time decouples computational dimension from hardware dimension. Exploiting space-time duality, we implement optical diffraction and interference entirely in time domain, using thin-film lithium niobate modulators as time lenses and temporal masks, with chromatic dispersion providing the coupling between successive temporal neurons. We experimentally verify high-order, complex-valued matrix-matrix multiplications using just a single optical input/output port, scaling the computational dimensions far beyond the channel count. By incorporating optical feedback, we extend this platform into versatile neural networks, where the network layers, neuron numbers, and synaptic connections are fully programmable and in-situ trainable. Our temporal diffractive neural networks are successfully validated on various classification benchmarks, alongside image and video generation tasks. Notably, using this platform we demonstrate an all-analogue generative pipeline in which the latent variable is drawn directly from amplified spontaneous emission, so that no digital sampling or electronic modulation appears anywhere in the generative path. Furthermore, high-resolution images and videos are generated at high frame rates, outperforming state-of-the-art modulator-refresh-limited optical generative systems. These results establish a unified photonic temporal computing framework, providing a scalable and deployable pathway toward next-generation machine intelligence.
ER -