A Compact Butterfly-Style Silicon Photonic–Electronic Neural Chip for Hardware-Efficient Deep Learning,ACS Photonics

当前位置： X-MOL 学术 › ACS Photonics › 论文详情

Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)

A Compact Butterfly-Style Silicon Photonic–Electronic Neural Chip for Hardware-Efficient Deep Learning
ACS Photonics ( IF 6.5 ) Pub Date : 2022-11-30 , DOI: 10.1021/acsphotonics.2c01188
Chenghao Feng _{1,

2} , Jiaqi Gu ₂ , Hanqing Zhu ₂ , Zhoufeng Ying _{1,

3} , Zheng Zhao _{2,

4} , David Z. Pan ₂ , Ray T. Chen _{1,

2,

5}

Affiliation

The optical neural network (ONN) is a promising hardware platform for next-generation neurocomputing due to its high parallelism, low latency, and low energy consumption. Previous ONN architectures are mainly designed for general matrix multiplication (GEMM), leading to unnecessarily large area cost and high control complexity. Here, we move beyond classical GEMM-based ONNs and propose an optical subspace neural network (OSNN) architecture, which trades the universality of weight representation for lower optical component usage, area cost, and energy consumption. We devise a butterfly-style photonic–electronic neural chip to implement our OSNN with up to 7× fewer trainable optical components compared to GEMM-based ONNs. Additionally, a hardware-aware training framework is provided to minimize the required device programming precision, lessen the chip area, and boost the noise robustness. We experimentally demonstrate the utility of our neural chip in practical image recognition tasks, showing that a measured accuracy of 94.16% can be achieved in handwritten digit recognition tasks with 3 bit weight programming precision.

中文翻译：

用于硬件高效深度学习的紧凑型蝶形硅光电子神经芯片

光学神经网络 (ONN) 具有高并行性、低延迟和低能耗等优点，是下一代神经计算的有前途的硬件平台。以前的 ONN 架构主要是为通用矩阵乘法 (GEMM) 而设计的，导致不必要的大面积成本和高控制复杂性。在这里，我们超越了经典的基于 GEMM 的 ONN，并提出了一种光学子空间神经网络 (OSNN) 架构，该架构以权重表示的普遍性为代价来降低光学组件的使用、面积成本和能耗。我们设计了一种蝴蝶式光子电子神经芯片来实现我们的 OSNN，与基于 GEMM 的 ONN 相比，可训练光学组件减少了 7 倍。此外，还提供了一个硬件感知培训框架，以最大限度地减少所需的设备编程精度，减小芯片面积，提高噪声鲁棒性。我们通过实验证明了我们的神经芯片在实际图像识别任务中的实用性，表明在 3 位权重编程精度的手写数字识别任务中可以实现 94.16% 的测量精度。

更新日期：2022-11-30

点击分享查看原文

点击收藏

公开下载

阅读更多本刊最新论文本刊介绍/投稿指南11