.png){kind=icon}
返回
Transferable learning on analog hardware
.png){kind=icon}
OA
AI
DOI：10.1126/sciadv.adh3436.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
While analog neural network (NN) accelerators promise massive energy and time savings, an important challenge is to make them robust to static fabrication error. Present-day training methods for programmable photonic interferometer circuits, a leading analog NN platform, do not produce networks that perform well in the presence of static hardware errors. Moreover, existing hardware error correction techniques either require individual retraining of every analog NN (which is impractical in an edge setting with millions of devices), place stringent demands on component quality, or introduce hardware overhead. We solve all three problems by introducing one-time error-aware training techniques that produce robust NNs that match the performance of ideal hardware and can be exactly transferred to arbitrary highly faulty photonic NNs with hardware errors up to five times larger than present-day fabrication tolerances.
Keyword:
HIGH-EXTINCTION-RATIO
NEURAL-NETWORKS
SWITCH
DESIGN
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
IF:
12.5
论文数:
2.1W
被引数:
18.1W
机构
暂无机构信息
引用论文
Setting up meshes of interferometers - reversed local light interference method.png)
OPTICS EXPRESSOAAI
OPTICS EXPRESSIF3.3
OAAI分享收藏
分享收藏Large-scale silicon quantum photonics implementing arbitrary two-qubit processing实现任意两量子位处理的大规模硅量子光子学NATURE PHOTONICSOAAI
NATURE PHOTONICSIF32.9
OAAI分享收藏A Complete Si Photonics Platform Embedding Ultra-Low Loss Waveguides for O- and C-Band一个完整的Si光子学平台，嵌入O和c波段的超低损耗波导OAAI
OAAI分享收藏Silicon Photonic Switch Subsystem With 900 Monolithically Integrated Calibration Photodiodes and 64-Fiber Package具有900单片集成校准光电二极管和64光纤封装的硅光子开关子系统
PREAI
分享收藏
查看更多内容