检索结果-内蒙古大学图书馆

Reservoir computing for vector-mode decomposition in deep learning frameworks

PHYSICA SCRIPTA 2025年第4期100卷 046014-046014页

作者： Li, Jian-Jun Yang, Feng Wu, Bao-Jian Wen, Feng Univ Elect Sci & Technol China Sch Informat & Commun Engn Key Lab Opt Fiber Sensing & Commun Minist Educ Chengdu 611731 Peoples R China Marolabs Co Ltd Lab Holog Opt Sensing Chengdu 610041 Peoples R China

This paper introduces an optical mode decomposition (OMD) framework for the core-shifted few-mode fiber (FMF) channel, leveraging the nonlinear mapping capabilities of reservoir computing (RC) integrated with deep learning architectures. The core-shift nature, characterized by radial and phase offsets, transforms the originally simpler OMD task-focused solely on mode weights and phases-into a significantly more complex problem. We integrate RC with deep learning, where RC replaces the nonlinear processing layer in traditional neural networks for hardware-friendly implementations. Simulation results demonstrate that the proposed network maintains a high correlation coefficient of 94% while reducing computational complexity by 32% for two vector modes. The framework's robustness is further validated for cases involving more vector modes, confirming its potential for practical hardware implementation. This study not only advances the field of OMD by addressing the complexities introduced by core shift but also provides a novel pathway for the hardware realization of deep learning models in optical communication systems.

关键词： optical mode decomposition reservoir computing deep learning few-mode fibre core shift

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised High-Order Information Bottleneck Learning of Spiking neural network for Robust Event-Based optical Flow Estimation

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2025年第4期47卷 2280-2297页

作者： Yang, Shuangming Linares-Barranco, Bernabe Wu, Yuzhu Chen, Badong Tianjin Univ Sch Elect & Informat Engn Tianjin 300072 Peoples R China CSIC Inst Microelect Sevilla Seville 41092 Spain Univ Seville Seville 41092 Spain Xi An Jiao Tong Univ Natl Engn Res Ctr Visual Informat & Applicat Natl Key Lab Human Machine Hybrid Augmented Intell Xian 710049 Peoples R China Xi An Jiao Tong Univ Inst Artificial Intelligence & Robot Xian 710049 Peoples R China

Event cameras form a fundamental foundation for visual perception in scenes characterized by high speed and a wide dynamic range. Although deep learning techniques have achieved remarkable success in estimating event-based optical flow, existing methods have not adequately addressed the significance of temporal information in capturing spatiotemporal features. Due to the dynamics of spiking neurons in SNNs, which preserve important information while forgetting redundant information over time, they are expected to outperform analog neural networks (ANNs) with the same architecture and size in sequential regression tasks. In addition, SNNs on neuromorphic hardware achieve advantages of extremely low power consumption. However, present SNN architectures encounter issues related to limited generalization and robustness during training, particularly in noisy scenes. To tackle these problems, this study introduces an innovative spike-based self-supervised learning algorithm known as SeLHIB, which leverages the information bottleneck theory. By utilizing event-based camera inputs, SeLHIB enables robust estimation of optical flow in the presence of noise. To the best of our knowledge, this is the first proposal of a self-supervised information bottleneck learning strategy based on SNNs. Furthermore, we develop spike-based self-supervised algorithms with nonlinear and high-order information bottleneck learning that employs nonlinear and high-order mutual information to enhance the extraction of relevant information and eliminate redundancy. We demonstrate that SeLHIB significantly enhances the generalization ability and robustness of optical flow estimation in various noise conditions. In terms of energy efficiency, SeLHIB achieves 90.44% and 45.70% cut down of energy consumption compared to its counterpart ANN and counterpart SNN models, while attaining 33.78% lower AEE (MVSEC), 5.96% lower RSAT (ECD) and 6.21% lower RSAT (HQF) compared to the counterpart ANN implementation

关键词： optical flow Estimation Neuromorphics Event detection Self-supervised learning Training Biomedical optical imaging Heuristic algorithms Robustness Neurons Spiking neural network optical flow estimation self-supervised learning event-based vision neuromorphic computing

来源：评论

学校读者我要写书评

暂无评论

Photonic Bayesian neural networks: Leveraging Programmable Noise for Robust and Uncertainty-Aware computing

引用

ADVANCED SCIENCE 2025年

作者： Zhuge, Yangyang Ren, Zhihao Xiao, Zian Zhang, Zixuan Liu, Xinmiao Liu, Weixin Xu, Siyu Ho, Chong Pei Li, Nanxi Lee, Chengkuo Natl Univ Singapore Dept Elect & Comp Engn 4 Engn Dr 3 Singapore 117583 Singapore Natl Univ Singapore Ctr Intelligent Sensors & MEMS 4 Engn Dr 3 Singapore 117583 Singapore ASTAR Inst Microelect IME 2 Fusionopolis Way Innovis 08-02 Singapore 138634 Singapore Natl Univ Singapore Natl Ctr Adv Integrated Photon NCAIP Singapore 119077 Singapore NUS Suzhou Res Inst NUSRI Suzhou 215123 Jiangsu Peoples R China Natl Univ Singapore NUS Grad Sch Integrat Sci & Engn Programme ISEP Singapore 119077 Singapore

Photonic neural networks (PNNs) based on silicon photonic integrated circuits (Si-PICs) offer significant advantages over microelectronic counterparts, including lower energy consumption, higher bandwidth, and faster computing speeds. However, the analog nature of optical signal in PNNs makes Si-PIC solutions highly sensitive to device noise, especially when using fixed-value deterministic models, which are not robust to hardware fluctuation. Furthermore, current PNNs are unable to handle data uncertainty, a critical factor in applications such as autonomous driving, medical diagnostics, and financial forecasting. Herein, a photonic Bayesian neural network (PBNN) architecture that incorporates Bayesian principles to enhance robustness and address uncertainty is proposed. In the PBNN, device noise is leveraged through photonic-noise-based random number generators, which combine Mach-Zehnder interferometers and micro-ring resonators to independently control output mean and standard deviation. Based on modelling with experimentally extracted data, the PBNN achieves a classification accuracy of up to 98% for handwritten digit recognition, matching full-precision models on conventional computers. Beyond classification, the PBNN excels in multimodal data processing, regression, and outlier detection. This scalable, energy-efficient architecture transforms photonic noise into computational value, addressing the limitations of deterministic PNNs and enabling uncertainty-aware computing for real-world applications.

关键词： Bayesian neural network multimodal data processing photonic probabilistic computing photonic random number generator silicon photonics

来源：评论

学校读者我要写书评

暂无评论

Fourier neural Operator Accelerated Fast and Accurate optical Fiber Transmission Modeling

引用

JOURNAL OF LIGHTWAVE TECHNOLOGY 2025年第8期43卷 3592-3604页

作者： Zhang, Tianhong Pan, Yutong Zheng, Jiayu Cai, Xiang Wang, Danshi Zhou, Xian Zhang, Fan Peking Univ Sch Elect State Key Lab Adv Opt Commun Syst & Networks Beijing 100871 Peoples R China Beijing Univ Posts & Telecommun State Key Lab Informat Photon & Opt Commun Beijing 100876 Peoples R China Univ Sci & Technol Beijing Sch Comp & Commun Engn Beijing Peoples R China Peng Cheng Lab Shenzhen 518055 Peoples R China

optical signal transmission in optical fiber links is governed by nonlinear Schr & ouml;dinger equation (NLSE). Fast and accurate modeling the propagation of optical signals is crucial for designing digital signal processing (DSP) algorithms, and optimizing the system performance. In this article, we combine Fourier neural operator (FNO) with the traditional numerical split-step Fourier method (SSFM) for solving NLSE, and propose a novel modeling scheme called simplified SSFM (S-SSFM) + FNO, which can perform fast modelling of long-distance transmission of high baud rate signals. The proposed method has been validated for accuracy, generalization, and rapid computational speed in both single-channel and wavelength-division-multiplexing (WDM) systems. In single-channel scenarios with transmission distances up to 2400 km, accounting for varying signal powers, the signal-to-noise ratio (SNR) errors of the proposed scheme remain consistently below 0.1dB. Remarkably, the calculation time for this method amounts to only 7.5% of the constant step size SSFM (C-SSFM) approach with a step size of 0.1 km, and 17.5% of the nonlinear phase SSFM (NP-SSFM) with a maximum phase shift value at 0.05 degree. Furthermore, the neural network (NN) utilized in this scheme demonstrates a notable advantage in training speed, achieving convergence in a timeframe comparable to the modeling phase. For WDM systems, the superior computational efficiency of S-SSFM+FNO becomes increasingly evident. With 21 channels, the computational complexity is reduced by an impressive 99.6% compared to C-SSFM and 98.5% compared to NP-SSFM, with the normalized mean squared error (NMSE) remains below 3.9e-3 for transmission over 800 km.

关键词： Computational modeling Mathematical models Accuracy Wavelength division multiplexing optical fiber dispersion Noise Training Nonlinear optics Fiber nonlinear optics Artificial neural networks Deep learning (DL) Fourier neural operator (FNO) optical fiber modeling split-step Fourier method (SSFM)

来源：评论

学校读者我要写书评

暂无评论

OMS-DNN: A Surrogate Model for Forward Design of Mirror Structure and optical System

引用

IEEE ACCESS 2025年 13卷 64853-64866页

作者： Zhang, Dongxue Deng, Xiaoqin Sha, Wei Wang, Jianzhao Lou, Hongwei Sheng, Lei Wang, Ye Zhu, Junqing Chinese Acad Sci Changchun Inst Opt Fine Mech & Phys Changchun 130033 Peoples R China

As demand grows for real-time diagnostics and control in mirror structure and optical system design, traditional computation-heavy methods face efficiency bottlenecks. To address this, we propose OMS-DNN, a fully data-driven deep learning framework that replaces resource-intensive finite element analysis and ray tracing, while integrating optimization algorithms for forward design. OMS-DNN models complex nonlinear relationships using deep neural networks and captures parameter topologies through a graph convolutional network (GCN). We evaluate OMS-DNN on large-scale datasets, including 20,000 structural and 100,000 Three-Mirror Anastigmat (TMA) optical design samples. It achieves mean absolute percentage errors (MAPE) of 0.98%, 4.09%, and 1.66% across three key metrics. To meet forward design requirements, we integrated the proposed architecture with an optimization algorithm. Experimental results demonstrate that our method can rapidly generate constraint-compliant parameter design solutions tailored to various needs, thereby lowering the entry barrier for designers. These findings validate the effectiveness of OMS-DNN and its potential to transform traditional design workflows.

关键词： optical design Nonlinear optics Mirrors Optimization optical imaging System analysis and design Feature extraction optical computing Adaptive optics Accuracy Deep learning mirror structure design optical system design

来源：评论

学校读者我要写书评

暂无评论

Enhancing quality inspection in automotive manufacturing through deep learning and transfer learning

引用

neural computing and Applications 2025年 1-26页

作者： El Wahabi, Abdelhamid Mesquiny, Ayoub El Ahmadi, Oumaima Baraka, Ibrahim Hadj Hamdoune, Salaheddine Boudhir, Anouar Abdelhakim LIST Laboratory ISEEC Research Team UAE Faculty of Science and Technology Tangier Morocco C3S Laboratory SSET Research Team UAE Faculty of Science and Technology Tangier Morocco

Ensuring quality control and accurate defect detection are paramount in the automotive manufacturing industry. Nevertheless, manual and automated optical inspection approaches have encountered limitations in efficiently addressing this area. To overcome these challenges, automotive companies are actively investigating the utilization of deep learning models to automate and optimize quality control processes. This paper focuses on the development of image classification models with the specific objective of accurately categorizing industrial components as acceptable or defective. The targeted manufacturing processes include terminal crimps, gluing the cap of coils, soldering pins and inductors. Our approach leverages transfer learning to mitigate the limited data, utilizing ten state-of-the-art convolutional neural network models (CNNs) previously developed for the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC): DenseNet121, DenseNet201, InceptionResNetV2, InceptionV3, MobileNetV2, ResNet50V2, ResNet101V2, ResNet152V2, VGG19, and Xception. The last layer of these models replaced to align with our dataset's classes, and various strategies such as freezing layers, fine-tuning, adding other layers, and retraining are employed. Finally, we evaluated our performances models using metrics such as accuracy, precision, recall, and ROC–AUC. The results demonstrate the efficiency of transfer learning in quality inspection classification, with DenseNet121 and VGG19 achieving accuracy and precision rates exceeding 99%. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Unveiling the relationship between Fabry-Perot laser structures and optical field distribution via symbolic regression

引用

Optoelectronics Letters 2025年第3期21卷 149-154页

作者： LI Wenqiang WU Min LI Weijun HAO Meilan YU Lina AnnLab Institute of SemiconductorsChinese Academy of SciencesBeijing100083China School of Electronic Electrical and Communication Engineering&School of Integrated CircuitsUniversity of Chinese Academy of SciencesBeijing100049China Beijing Key Laboratory of Semiconductor Neural Network Intelligent Sensing and Computing Technology Beijing100083China

In recent years,machine learning(ML)techniques have been shown to be effective in accelerating the development process of optoelectronic ***,as"black box"models,they have limited theoretical *** this work,we leverage symbolic regression(SR)technique for discovering the explicit symbolic relationship between the structure of the optoelectronic Fabry-Perot(FP)laser and its optical field distribution,which greatly improves model transparency compared to *** demonstrated that the expressions explored through SR exhibit lower errors on the test set compared to ML models,which suggests that the expressions have better fitting and generalization capabilities.

关键词： machine learning optoelectronic deviceshoweverasblack optical field distributionwhich symbolic regression symbolic regression sr technique Fabry Perot laser discovering explicit symbolic relationship optical field distribution

来源：评论

学校读者我要写书评

暂无评论

AGBUNet: an enhanced CNN-UNET architecture for the prediction of above ground biomass using deep learning

引用

neural computing and Applications 2025年第5期37卷 3809-3826页

作者： Arumai Shiney, S. Geetha, R. Department of Computer Science and Engineering S.A. Engineering College Chennai India

Accurate prediction of above ground biomass (AGB) is critical for monitoring forest health and carbon cycling. It is crucial for understanding and managing forest ecosystems. In this paper, we propose an enhanced framework combining convolutional neural network (CNN) and UNet, termed AGBUNet, specifically designed for predicting AGB using remote sensing data. The framework consists of separate CNNs for processing each type of image, whose outputs are subsequently fused in a UNet architecture to enhance prediction accuracy. These modifications include customized convolutional layers, advanced preprocessing techniques, and a novel integration of data prompts from separate Sentinel-1 and Sentinel-2 image processing streams. The AGBUNet integrates Sentinel-1 and Sentinel-2 images to leverage complementary information from synthetic aperture radar (SAR) and optical sensors. This study underscores the potential of the AGBUNet model for enhancing biomass estimation from remote sensing data, contributing to better forest management and ecological monitoring. The performance measures obtained are compared with the other models, and the following results are obtained as follows for MSE value of 298.25, RMSE value of 15.27 and MAE value of 12.21, and the values are satisfying compared with earlier benchmarks. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Building Scalable Silicon Microring Resonator-Based Neuromorphic Photonic Circuits Using Post-Fabrication Processing with Photochromic Material

引用

ADVANCED optical MATERIALS 2025年第11期13卷

作者： Xu, Lei Zhang, Jiawei Doris, Eli A. Bilodeau, Simon Wisch, Jesse A. Gui, Manting Jimoh, Yusuf O. Shastri, Bhavin Rand, Barry P. Prucnal, Paul R. Princeton Univ Dept Elect & Comp Engn 41 Olden St Princeton NJ 08544 USA Queens Univ Dept Phys Engn Phys & Astron Kingston ON K7L 3N6 Canada Vector Inst Toronto ON M5G 1M1 Canada

Neuromorphic photonics has become one of the research forefronts in photonics, with its benefits in low-latency signal processing and potential in significant energy consumption reduction when compared with digital electronics. With artificial intelligence (AI) computing accelerators in high demand, one of the high-impact research goals is to build scalable neuromorphic photonic integrated circuits which can accelerate the computing of AI models at high energy efficiency. A complete neuromorphic photonic computing system comprises seven stacks: materials, devices, circuits, microarchitecture, system architecture, algorithms, and applications. Here, we consider microring resonator (MRR)-based network designs toward building scalable silicon integrated photonic neural networks (PNN), and variations of MRR resonance wavelength from the fabrication process and their impact on PNN scalability. Further, post-fabrication processing using organic photochromic layers over the silicon platform is shown to be effective for trimming MRR resonance wavelength variation, which can significantly reduce energy consumption from the MRR-based PNN configuration. Post-fabrication processing with photochromic materials to compensate for the variation in MRR fabrication will allow a scalable silicon system on a chip without sacrificing today's performance metrics, which will be critical for the commercial viability and volume production of large-scale silicon photonic circuits.

关键词： neuromorphic photonics photonic neural networks photochromic material silicon photonics

来源：评论

学校读者我要写书评

暂无评论

Mechanical properties estimation of multi-layer friction stir plug welded aluminium plates using time-series neural network models

引用

Soft computing 2025年第2期29卷 1147-1168页

作者： Qazani, Mohammad Reza Chalak Sajed, Moosa Pedrammehr, Siamak Seyedkashi, Seyed Mohammad Hossein Sohar University Sohar311 Oman Mechanical Engineering Department Azarbaijan Shahid Madani University Tabriz Iran Faculty of Design Tabriz Islamic Art University Tabriz Iran Mechanical Engineering Department University of Birjand Birjand97175-376 Iran

A multi-layer friction stir plug welding can be used to fix the thick aluminium plates. optical microscopy and tensile tests are utilized to study the microstructural and mechanical characteristics of the welded aluminium plates. However, finding the relation between the indexes of the process and the mechanical properties would be challenging. The present work aims to devise a time-series machine learning model including a recurrent neural network (RNN) and nonlinear autoregressive network with the external state (NARX) to estimate the mechanical properties of the repaired aluminium plate using the force-extension plot. The ultimate tensile strength, yield strength, impact energy and elongation of the repaired aluminium plate can be calculated based on a force-extension plot trained and extracted using the developed networks. In addition, the Bayesian technique is employed to recalculate the optimal hyperparameters of RNN and NARX, targeting the lowest root mean square error (RMSE) between the target and the estimated force during the testing. The investigated methods (RNN and NARX) with the addition of classical estimation methods, including decision tree and support vector regression, are modelled in MATLAB, and the outcomes prove the proposed NARX model efficiency in terms of lower RMSE in comparison with support vector regression, decision tree and RNN. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Tensile testing

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：