RESEARCH ARTICLE

A fully real-valued end-to-end optical neural network for generative model    

Shan Jiang, Bo Wu, Qixiang Cheng, Jianji Dong

2026, 19(1): 4.https://doi.org/10.2738/foe.2026.0004

Optical neural networks (ONNs) hold great promise for low-latency, energy-efficient inference. However, the absence of a fully real-valued end-to-end ONN, in which the inputs, weight matrices, and nonlinear activations are all represented in the real-number domain and can be optically cascaded, remains a key bottleneck. Existing approaches either rely on electrical post-processing of photodetector outputs to extend the number field in the linear layers, which breaks optical cascadability, or employ photodiode–driven micro-ring modulators (MRMs) to implement nonlinearities, constraining subsequent-layer inputs to the nonnegative domain and thereby limiting network expressivity and architectural flexibility. Here, we employ two MRMs biased at different resonance wavelengths to achieve real-valued optical encoding, together with a dual-MRM activation element driven by the differential photocurrent of photodiodes, which provides optically cascadable real-valued nonlinear activation. Combined with a real-valued Mach–Zehnder interferometer mesh for matrix computation, this architecture realizes a fully real-valued end-to-end ONN. We experimentally demonstrate a tanh-like nonlinear activation function and validate it on an iris classification task, achieving an accuracy of 98%. We further model the generator of a generative adversarial network based on this structure, in which the nonlinear activation is based on the experimentally measured nonlinear transfer curve. The generator can use natural optical noise as its input, thereby eliminating electro-optic conversion and digital-to-analog conversion at the input stage. With the above merits, the proposed ONN achieves successful optical-to-optical on-chip image generation, validating the superiority of optical computing.

Shan Jiang, Bo Wu, Qixiang Cheng, Jianji Dong. A fully real-valued end-to-end optical neural network for generative model. Front. Optoelectron., 2026, 19(1): 4 https://doi.org/10.2738/foe.2026.0004

研究背景

光学神经网络（ONN）具有低延迟、高能效等优势，有望用于下一代人工智能加速器。然而，现有ONN架构大多无法实现全实值、端到端的光学计算——即输入、权重矩阵和非线性激活函数均在实数域内，且能够光学级联。现有方法要么依赖电学后处理（破坏光学级联性），要么将非线性激活限制在非负域（如ReLU），从而损失了表达能力，对于生成对抗网络（GAN）等需要零均值、对称激活的任务尤其不利。

主要内容

本文提出并实验演示了一种全实值、端到端的光学神经网络。该架构利用两个偏置在不同谐振波长的微环调制器（MRM）实现实值光学编码，利用由差分光电流驱动的双MRM激活单元实现可光学级联的实值非线性激活（类tanh函数）。将此激活单元与可实现实值矩阵运算的MZI网格相结合，构建了完整的光学芯片。实验验证了该芯片在鸢尾花分类任务上的收敛性，并基于实测非线性曲线仿真了GAN生成器，利用部分相干光的自然噪声作为输入，成功实现了光域图像生成。

创新点

l 全实值端到端ONN架构：同时实现了实值输入编码、实值权重矩阵和实值非线性激活，且保持光学可级联性，突破了以往非负域限制。

l 新型差分光电流驱动的双MRM激活单元：通过将差分光电流注入一对并联但极性相反的MRM，实现了对正负输入均可响应的类tanh光学非线性激活函数，其输入输出映射覆盖第二和第四象限。

l 利用部分相干光噪声作为生成模型的输入：在GAN生成器中，采用滤波后的放大自发辐射（ASE）噪声作为随机输入，免去了电学伪随机数生成和数模转换。

l 高效波长复用与资源节约：所有层可共享同一套波长（正/负两组），所需波长通道数仅取决于网络宽度而非深度，且通过部分相干光可进一步减少至2个波长。

结论

本研究成功实现了首个全实值、端到端的光学神经网络，通过双MRM编码和差分光电流驱动的双MRM激活，突破了传统ONN只能处理非负信号的限制。实验验证了其在分类任务中的高精度（98%），并展示了在GAN生成模型中，实值域操作能够显著提升图像生成质量。此外，利用部分相干光作为片上随机噪声源，进一步降低了系统对电学接口的依赖。该工作为构建更强大、更灵活的光学计算平台，特别是面向生成式人工智能的光学加速器，奠定了坚实的基础。

（以上文字包含AI生成内容，仅供参考；请以原文为准。）