检索结果-内蒙古大学图书馆

International Symposium on Microwave, Antenna, Propagation and EMC Technologies for Wireless Communications (MAPE)

作者： Dingcheng Yi Zhiguo Zhang Wei Zhang Haitian Hu Haojie Zhang Beijing University of Posts and Telecommunications Beijing China

ISBN: (数字)9798331533922

ISBN: (纸本)9798331533939

One of the critical performance metrics of anti-resonant fibers(ARFs) is the confinement loss of the fundamental mode, and achieving a lower confinement loss is one of the primary objectives in current design endeavors. Traditionally, the process of obtaining the confinement loss of designed fibers through simulation software like COMSOL is excessively time-intensive. To expedite the modeling process of anti-resonant fibers, this paper explores how to leverage a well-trained artificial neural network for rapid simulation of anti-resonant fibers while achieving high-precision output predictions. In the proposed method, an artificial neural network model is established to effectively map the feature space to the target optical characteristics. The effectiveness of this method is demonstrated using a specific structure of anti-resonant fiber as an example.

关键词： optical fiber losses Microwave antennas Wireless communication optical propagation Artificial neural networks optical computing optical fiber networks Predictive models Propagation losses Software

来源：评论

学校读者我要写书评

暂无评论

High-throughput optical neural networks based on temporal computing

arXiv

引用

arXiv 2023年

作者： Zheng, Shuang Zhang, Jiawei Yu, Weizhen Zhang, Weifeng Radar Research Lab School of Information and Electronics Beijing Institute of Technology Beijing100081 China Ministry of Education Beijing100081 China Chongqing Innovation Centre Beijing Institute of Technology Chongqing401120 China Chongqing Key Laboratory of Novel Civilian Radar Chongqing401120 China

An emerging generative artificial intelligence (AI) based on neural networks starts to grow in popularity with a revolutionizing capability of creating new and original content. As giant generative models with millions to billions of parameters are developed, trained and maintained, a massive and energy-efficient computing power is highly required. However, conventional digital computers are struggling to keep up with the pace of the generative model improvements. In this paper, we propose and demonstrate high-throughput optical neural networks based on temporal computing. The core weighted summation operation is realized with the use of high-speed electro-optic modulation and low-speed balanced photodetection. The input data and weight are encoded in a time sequence separately and loaded on an optical signal via two electro-optic modulators sequentially. By precisely controlling the synchronization time of the data and weight loading, the matrix multiplication is performed. Followed by a balanced photodetector, the summation is conducted, thanks to the electron accumulation of the inherent electronic integrator circuit of the low-speed photodetector. Thus, the linear weighted summation operation is implemented based on temporal computing in the optical domain. With the proposed optical linear weighted summation, a fully-connected neural network and convolutional neural network are realized. Thanks to the high-speed feature of temporal computing, a high data throughput of the optical neural network is experimentally demonstrated, and the weighting coefficients can be specified on demand, which enables a strong programmability of the optical neural network. By leveraging wavelength multiplexing technology, a scalable optical neural network could be created with a massive computing power and strong reconfigurability, which holds great potential for future giant AI applications. Copyright © 2023, The Authors. All rights reserved.

关键词： Photodetectors

来源：评论

学校读者我要写书评

暂无评论

NeurOLight: A Physics-Agnostic neural Operator Enabling Parametric Photonic Device Simulation 36

NeurOLight: A Physics-Agnostic Neural Operator Enabling Para...

引用

36th Conference on neural Information Processing Systems (NeurIPS)

作者： Gu, Jiaqi Gao, Zhengqi Feng, Chenghao Zhu, Hanqing Chen, Ray T. Boning, Duane S. Pan, David Z. Univ Texas Austin Austin TX 78712 USA MIT Cambridge MA 02139 USA

ISBN: (纸本)9781713871088

optical computing is an emerging technology for next-generation efficient artificial intelligence (AI) due to its ultra-high speed and efficiency. Electromagnetic field simulation is critical to the design, optimization, and validation of photonic devices and circuits. However, costly numerical simulation significantly hinders the scalability and turn-around time in the photonic circuit design loop. Recently, physics-informed neural networks have been proposed to predict the optical field solution of a single instance of a partial differential equation (PDE) with pre-defined parameters. Their complicated PDE formulation and lack of efficient parametrization mechanisms limit their flexibility and generalization in practical simulation scenarios. In this work, for the first time, a physics-agnostic neural operator-based framework, dubbed NeurOLight, is proposed to learn a family of frequency-domain Maxwell PDEs for ultra-fast parametric photonic device simulation. We balance the efficiency and generalization of NeurOLight via several novel techniques. Specifically, we discretize different devices into a unified domain, represent parametric PDEs with a compact wave prior, and encode the incident light via masked source modeling. We design our model with parameter-efficient cross-shaped NeurOLight blocks and adopt superposition-based augmentation for data-efficient learning. With these synergistic approaches, NeurOLight generalizes to a large space of unseen simulation settings, demonstrates 2-orders-of-magnitude faster simulation speed than numerical solvers, and outperforms prior neural network models by similar to 54% lower prediction error with similar to 44% fewer parameters. Our code is available at link.

关键词： Photonic devices

来源：评论

学校读者我要写书评

暂无评论

Human Action Recognition Based on Three-Stream network with Frame Sequence Features 7

Human Action Recognition Based on Three-Stream Network with ...

引用

7th International Conference on Image, Vision and computing, ICIVC 2022

作者： Huang, Ruifeng Chen, Chong Cheng, Rui Zhang, Ying Zhu, Jiabing Anhui Jianzhu University Hefei230601 China Intellisencing and Multi-Dimensional Modeling Anhui International Joint Research Center for Ancient Architecture Hefei230601 China Huainan Normal University Huainan232038 China

ISBN: (数字)9781665467346

ISBN: (纸本)9781665467346

In the field of human action recognition (HAR), two-stream models have been widely employed. In recent years, traditional two-stream network models have disregarded the interframe sequence characteristics of video, resulting in a decrease in model robustness when local sequence information and long-term motion information interact. In light of this, a novel three-stream neural network is proposed by combining the long-term and short-term characteristics of a frame sequence with spatio-temporal information. Initially, the optical flow sequence image frames and RGB image frames in the video are extracted, the optical flow motion information and image space information in the video is obtained, the corresponding time network and space network are entered, and the spatial information is entered into the sequence feature processing network;the three networks are then pretrained. At the conclusion of training, the operation of feature extraction is executed, the features are incorporated with the parallel fusion algorithm by adding weights, and the behavior categories are classified using Multi-Layer Perception. Experimental results on the UCF11, UCF50, and HMDB51 datasets demonstrate that our model effectively integrates the spatial-temporal and frame-sequence information of human actions, resulting in a significant improvement in recognition accuracy. Its classification accuracy on the three datasets was 99.17%, 97.40%, and 96.88%, respectively, significantly enhancing the generalization capability and validity of conventional two-stream or three-stream models. © 2022 IEEE.

关键词： optical flows

来源：评论

学校读者我要写书评

暂无评论

Research on CTR Prediction Model Based on Deep neural network 2

Research on CTR Prediction Model Based on Deep Neural Networ...

引用

2nd International Conference on Artificial Intelligence, Automation, and High-Performance computing, AIAHPC 2022

作者： Zhang, Anfeng Qi, Guanglei Century College Beijing University of Posts and Telecommunications Beijing102101 China

ISBN: (纸本)9781510657717

With the advent of the era of big data, it takes a lot of time and manpower to build a model that can automatically mine the effective features of data and get the user click-through rate prediction model through training, because the singlemodel classifier needs to extract effective features and input them into the model training by means of feature processing and data mining. In view of the shortcomings of website click-through rate prediction technology, this paper proposes a model based on compressed incentive network to extract the influence of a single factor in the overall project characteristics. The paper also applies the neuron mechanism to extract the effective features of the original features, fuses the features proposed by the two models into new features and adds them to the deep neural network for training. Experiments are designed to prove the rationality of the method. The experimental results show that compared with the current commonly used model, it can improve the AUC index and keep the efficiency within a reasonable range. The research results of this paper are not only of great significance to the development of science and technology, but also affect our daily life and economic consumption all the time. © 2022 SPIE. All rights reserved.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Graph neural networks for Interpretable Tactile Sensing 27

Graph Neural Networks for Interpretable Tactile Sensing

引用

27th International Conference on Automation and computing, ICAC 2022

作者： Fan, Wen Bo, Hongbo Lin, Yijiong Xing, Yifan Liu, Weiru Lepora, Nathan Zhang, Dandan Dept. of Engineer Mathematics University of Bristol Bristol United Kingdom Dept. of Computer Science University of Bristol Bristol United Kingdom

ISBN: (数字)9781665498074

ISBN: (纸本)9781665498074

Fine-grained tactile perception of objects is significant for robots to explore the unstructured environment. Recent years have seen the success of Convolutional neural networks (CNNs)-based methods for tactile perception using high-resolution optical tactile sensors. However, CNNs-based approaches may not be efficient for processing tactile image data and have limited interpretability. To this end, we propose a Graph neural network (GNN)-based approach for tactile recognition using a soft biomimetic optical tactile sensor. The obtained tactile images can be transformed into graphs, while GNN can be used to analyse the implicit tactile information among the tactile graphs. The experimental results indicate that with the proposed GNN-based method, the maximum tactile recognition accuracy can reach 99.53%. In addition, Gradient-weighted Class Activation Mapping (Grad-CAM) and Unsigned Grad-CAM (UGrad-CAM) methods are used for visual explanations of the models. Compared to traditional CNNs, we demonstrated that the generated features of the GNN-based model are more intuitive and interpretable. © 2022 IEEE.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

SafeLight: Enhancing Security in optical Convolutional neural network Accelerators

arXiv

引用

arXiv 2024年

作者： Afifi, Salma Thakkar, Ishan Pasricha, Sudeep Department of Electrical and Computer Engineering Colorado State University Fort Collins United States Department of Electrical and Computer Engineering University of Kentucky Lexington United States

The rapid proliferation of deep learning has revolutionized computing hardware, driving innovations to improve computationally expensive multiply-and-accumulate operations in deep neural networks. Among these innovations are integrated silicon-photonic systems that have emerged as energy-efficient platforms capable of achieving light speed computation and communication, positioning optical neural network (ONN) platforms as a transformative technology for accelerating deep learning models such as convolutional neural networks (CNNs). However, the increasing complexity of optical hardware introduces new vulnerabilities, notably the risk of hardware trojan (HT) attacks. Despite the growing interest in ONN platforms, little attention has been given to how HT-induced threats can compromise performance and security. This paper presents an in-depth analysis of the impact of such attacks on the performance of CNN models accelerated by ONN accelerators. Specifically, we show how HTs can compromise microring resonators (MRs) in a state-of-the-art non-coherent ONN accelerator and reduce classification accuracy across CNN models by up to 7.49% to 80.46% by just targeting 10% of MRs. We then propose techniques to enhance ONN accelerator robustness against these attacks and show how the best techniques can effectively recover the accuracy drops. © 2024, CC BY-NC-ND.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Ship Detection from Satellite Imagery Using RetinaNet with Instance Segmentation 11th

Ship Detection from Satellite Imagery Using RetinaNet with I...

引用

11th International Conference on Recent Trends in computing, ICRTC 2022

作者： Dhorajiya, Arya Rakhi, Anusree Mondal Saranya, P. Department of Computing Technologies SRM Institute of Science and Technology Kattankulathur Chennai India

ISBN: (纸本)9789811988240

Recognizing the centrality of ship detection and segmentation based on optical remote sensing images in the military and civilian domains, the sea surface remote sensing field is receiving a lot of attention. The deep learning-based ship detection approach can also extract vessel position and category information, crucial for maritime surveillance. Unfortunately, less research has been done in this area. The majority of known approaches for spotting ships perform object detection but don't perform semantic segmentation. Furthermore, earlier studies included limitations such as the inability to detect small ships or ships of diverse sizes and the higher rate of false positives. Convolutional neural networks (CNNs) benefits are also significantly reduced due to the complexities of SAR images. Then there are a few models that can't discern the difference between items that seem like ships, and these false alarms lower the total accuracy rate. The available models have also faced trouble while identifying and segmenting ships with a complicated background and dealing with light variations. Furthermore, some existing models performed somewhat worse than other state-of-the-art frameworks. Some models have a higher cost of computation. This paper uses RetinaNet to provide the ship detecting algorithm. Instance Segmentation was also included in the model. The proposed model was created by integrating these two strategies. RetinaNet is a fantastic one-stage object detection model that can precisely work with dense and small-scale objects. The Feature Pyramid network (FPN) and Focal Loss are two improvements over conventional one-step object detection models. Then basically, an instance segmentation algorithm has been implemented in place of the object detector to increase RetinaNet's performance by removing background better. The proposed model can work more precisely than existing models due to these adjustments. We achieved a promising result using the described techniques

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

The Character Recognition Method Based on OCR 26

The Character Recognition Method Based on OCR

引用

26th ACIS International Winter Conference on Software Engineering, Artificial Intelligence, networking and Parallel/Distributed computing, SNPD-Winter 2023

作者： Jianyang, Liu Junrong, Bao Bingjin, Li Zhiang, Feng Su, Zhang Yunnan Province Information Technology Development Center Yunnan China

ISBN: (纸本)9798350345865

The application scenarios of using digital documents for information exchange in modern society are gradually increasing with the rapid development of new technologies. Manually completing the classification and entry of book texts, book lending certificates and other information will also consume a lot of manpower, material resources and time costs. OCR technology based on deep learning processes data efficiently and with high accuracy which can help users better manage and process information. Therefore, this paper proposes an optical character recognition method based on deep neural network, integrating the DB Net model and the CRNN model. The main work is as follows: (1) We investigate and compare different deep learning-based text detection and text recognition algorithms. (2) Based on the features of random text position and directional orientation in text images under natural scenes, the text recognition method of DB Net and CRNN with CTC is proposed. DB Net Model outperforms other text detection models in terms of accuracy and speed and the accuracy of CRNN model can meet the requirements of real business scenarios. (3) We use the dataset to train the algorithm model and expand the method structure combined with the front-end framework. The method meets the task requirements of users to identify images and text accurately and provides a reference value for the study of the same type of methods. © 2023 IEEE.

关键词： optical character recognition

来源：评论

学校读者我要写书评

暂无评论

Model authentication hashing: Identifying pirated neural network models

引用

Expert Systems with Applications 2025年 283卷

作者： Xiong, Cheng Feng, Guorui Sun, Yunlong Qian, Zhenxing Yao, Heng Qin, Chuan School of Optical-Electrical and Computer Engineering University of Shanghai for Science and Technology Shanghai200093 China School of Communication and Information Engineering Shanghai University Shanghai200444 China Department of Tourism Fudan university Shanghai200433 China Key Laboratory of Culture & Tourism Intelligent Computing of China Ministry of Culture and Tourism Shanghai200433 China School of Computer Science Fudan University Shanghai200433 China

Nowadays, a great number of neural network models have been trained and applied in many fields. At the same time, the emergence of vast network models has already highlighted the security problems of intellectual property protection. High-value pre-trained models of original inventors may be pirated and used by illegal users without authorization. In order to solve this problem, we first give the definition of model differentiation, and then propose an effective robust hashing method based on model parameters for piracy identification in this paper. Through our model hash generation network, the parameters of original model are transformed into a one-way hash sequence with the fixed length, and the robustness and discrimination of the model hashes can be utilized to authenticate model parameters. Experimental results show that the accuracy rates of similar model detection and different model discrimination under different cases are almost all greater than 96.5%, and the comprehensive performance of proposed method is also the best one compared with other 3 SOTA methods. These results mean that the proposed method can be applied for identifying illegal model piracy effectively. © 2025 Elsevier Ltd

关键词： Terrorism

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：