检索结果-内蒙古大学图书馆

Enabling distributed Inference of Large neural networks on Resource Constrained Edge Devices using Ad Hoc networks 1

引用

20th International Symposium on distributed Computing and Artificial Intelligence, DCAI 2023

作者： Ohlenforst, Torsten Schreiber, Moritz Kreyß, Felix Schrauth, Manuel Am Wolfsmantel 33 Erlangen91058 Germany

ISBN: (数字)9783031383335

ISBN: (纸本)9783031383328

processing neural network inferences on edge devices, such as smartphones, IoT devices and smart sensors can provide substantial advantages compared to traditional cloud-based computation. These include technical aspects such as low latency or high data throughput – but also data sovereignty can be a concern in many applications. Even though general approaches of distributed inference have been developed recently, a transfer of these principles to the edge of the network is still missing. In an extreme edge setting, computations typically are severely constraint by available energy resources and communication limitations among individual devices, which makes the distribution and execution of large-scale neural networks particularly challenging. Moreover, since mobile devices are volatile, existing static networks are unsuited and instead a highly dynamic network architecture needs to be orchestrated and managed. We present a novel, multi-stage concept which tackles all associated tasks in one framework. Specifically, distributed inference approaches are complemented with necessary resource management and network orchestration in order to enable distributed inference in the field – hence paving the way towards a broad number of possible applications, such as autonomous driving, traffic optimization, medical applications, agriculture, and industry 4.0. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Energy resources

来源：评论

学校读者我要写书评

暂无评论

Road network Representation Learning with the Third Law of Geography 38

Road Network Representation Learning with the Third Law of G...

引用

38th Conference on neural Information processing Systems, NeurIPS 2024

作者： Zhou, Haicang Huang, Weiming Chen, Yile He, Tiantian Cong, Gao Ong, Yew-Soon College of Computing and Data Science Nanyang Technological University Singapore Institute of High-Performance Computing Agency for Science Technology and Research Singapore Centre for Frontier Al Research Agency for Science Technology and Research Singapore

Road network representation learning aims to learn compressed and effective vectorized representations for road segments that are applicable to numerous tasks. In this paper, we identify the limitations of existing methods, particularly their overemphasis on the distance effect as outlined in the First Law of Geography. In response, we propose to endow road network representation with the principles of the recent Third Law of Geography. To this end, we propose a novel graph contrastive learning framework that employs geographic configuration-aware graph augmentation and spectral negative sampling, ensuring that road segments with similar geographic configurations yield similar representations, and vice versa, aligning with the principles stated in the Third Law. The framework further fuses the Third Law with the First Law through a dual contrastive learning objective to effectively balance the implications of both laws. We evaluate our framework on two real-world datasets across three downstream tasks. The results show that the integration of the Third Law significantly improves the performance of road segment representations in downstream tasks. Our code is available at https://***/Haicang/Garner. © 2024 neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Fractional-order bi-Hopfield neuron coupled via a multistable memristor: Complex neuronal dynamic analysis and implementation with microcontroller

引用

AEU-INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATIONS 2025年 191卷

作者： Tamba, Victor Kamdoum Biamou, Arsene Loic Mbanda Pham, Viet-Thanh Grassi, Giuseppe Tagne, Francois Kapche Takougang, Armand Cyrille Nzeukou Univ Dschang Dept Telecommun & Network Engn IUT Fotso Victor Bandjoun POB 134 Bandjoun Cameroon Univ Dschang Dept Elect Engn Res Unit Automat & Appl Comp IUT FV Bandjoun POB 134 Bandjoun Cameroon Ind Univ Ho Chi Minh City Fac Elect Technol Ho Chi Minh City Vietnam Univ Salento Dept Engn Innovat I-73100 Lecce Italy Univ Dschang Fac Sci Dept Phys Res Unit Condensed Matter Elect & Signal Proc POB 67 Dschang Cameroon

The Hopfield neuron is an artificial neuron model used for pattern memorization and recognition. It exhibits a complex dynamic with stable states corresponding to memorized patterns. In order to grasp a more complete representation of the information exchange between two neurons, emphasizing the importance of neuronal connections in brain processing, we propose in this work the coupling of two fractional-order Hopfield neurons via a fractional-order flux-controlled multistable memristor. Each of these two neurons incorporates a selfcoupling memristive component, called an autapse memristor. Additionally, the second neuron is subjected to an external electromagnetic radiation, simulated by an additional memristor. The I-V characteristics of the memristors integrated in this model are analyzed through numerical simulations. The simulations of model dynamics versus its diverse parameters have revealed rich and complex dynamical behaviors. These simulations demonstrate that the proposed model generates a variety of homogeneous and heterogeneous chaotic attractors, distributed at diverse locations. The elaborated memristor coupled fractional-order bi-Hopfield neuron MCFBHN model is implemented on an Arduino-Due platform. A comparison of the results of the two approaches shows good consistency.

关键词： Fractional-order memristive Hopfield neural network Autapse memristor External electromagnetic radiation memristor Memristor initial boosting behaviors Microcontroller implementation

来源：评论

学校读者我要写书评

暂无评论

Analog Two-Variable Spiking Neuron in 16 nm FinFET for Neuromorphic Systems 19

Analog Two-Variable Spiking Neuron in 16 nm FinFET for Neuro...

引用

19th Conference on Ph.D Research in Microelectronics and Electronics (PRIME)

作者： Stevenazzi, Lorenzo Baschirotto, Andrea De Matteis, Marcello Univ Milano Bicocca Dept Phys Piazza Sci 3 I-20126 Milan Italy

ISBN: (纸本)9798350386301;9798350386318

The deployment of machine learning algorithms on edge devices is challenging due to the required low power consumption and high computing power. Neuromorphic systems operate in a parallel and distributed way, with multiple processing elements that mimic or simulate biological neurons on silicon. This approach is promising in equipping sensors with alwayson signal processing capabilities, naturally executing spiking neural network algorithms on dedicated hardware. Analog silicon neurons exhibit the lowest energy per spike, but are traditionally implemented in planar technology processes. In order to increase their density, hence the computing power, in the same area, it is essential to explore implementations in more scaled technology nodes. In this work an analog two-variable spiking neuron is described and implemented in 16 nm FinFET, adopting a supply voltage of 400 mV and employing transistors in the subthreshold region to reduce power consumption. Transient simulations for the regular spiking and fast spiking patterns are reported, showing the accelerated timescale at which the spiking neuron operates, with 38 mu s and 7 mu s time interval between two consecutive spikes respectively. With a required energy per spike of 117.5 fJ for the regular spiking configuration and the possibility of tuning external voltage references to modify the resulting spiking pattern, the presented circuit is suitable for developing large-scale neuromorphic architectures.

关键词： Neuromorphic Hardware Spiking Neuron Analog Circuit FinFET neural network

来源：评论

学校读者我要写书评

暂无评论

Synchronization of Multi-Term Fractional-Order neural networks with Switching Parameters via Hybrid Impulsive Control 12

Synchronization of Multi-Term Fractional-Order Neural Networ...

引用

12th IFAC Conference on Fractional Differentiation and its Application (ICFDA)

作者： Yang, Dongsheng Wang, Hu Zhang, Xiaoli Yu, Yongguang Beijing Jiaotong Univ Sch Math & Stat Beijing 100044 Peoples R China Cent Univ Finance & Econ Sch Math & Stat Beijing 100081 Peoples R China

This paper investigates the hybrid impulsive control synchronization problem of multi-term fractional-order neural networks (MFNNs) with switching parameters. Firstly, a novel MFNN with switching parameters model is introduced by incorporating multi-term Caputo fractional-order derivative to extend the existing framework for fractional-order cases. Then, the relationship between multi-term fractional-order derivative and distributed-order derivative is analyzed, and a synchronization criterion for a class of multi-term fractional-order impulsive switched systems is derived by utilizing the properties of the distributed-order derivative weight function. Furthermore, a hybrid impulsive controller is designed to obtain sufficient conditions for synchronization of MFNNs with switching parameters. To validate the effectiveness of the obtained conclusions, a numerical example is presented to demonstrate the validity of the proposed MFNN model and synchronization criterion. Copyright (C) 2024 The Authors. This is an open access article under the CC BY-NC-ND license (https://***/licenses/by-nc-nd/4.0/)

关键词： drive-response synchronization multi-term fractional-order switched neural network impulsive control

来源：评论

学校读者我要写书评

暂无评论

Semi-Passive RIS-Aided Sequential Channel Estimation and Prediction 13

Semi-Passive RIS-Aided Sequential Channel Estimation and Pre...

引用

13rd IEEE Sensor Array and Multichannel Signal processing Workshop (SAM)

作者： Haider, Mirza A. Zhang, Yimin D. Ding, Yanwu Shen, Dan Khanh Pham Chen, Genshe Temple Univ Dept Elect & Comp Engn Philadelphia PA 19122 USA Wichita State Univ Dept Elect & Comp Engn Wichita KS 67260 USA Intelligent Fus Technol Inc Germantown MD 20876 USA Air Force Res Lab Kirtland AFB NM 87117 USA

ISBN: (纸本)9798350344820;9798350344813

The popularity of mmWave in 5G and future communications is hindered by challenging propagation environments, such as line-of-sight obstruction. Reconfigurable intelligent surfaces (RIS) address this issue by dynamically modifying wireless channels, thereby enhancing data rates, reducing latency, and improving reliability in non-line-of-sight scenarios. For high data-rate communication and precise mobile user localization, a large RIS is required, resulting in a high pilot overhead for channel estimation. To address this issue, we exploit a semi-passive RIS with sparsely distributed active RIS elements in lieu of fully passive RIS. This approach efficiently enables channel estimation both between the base station and the RIS as well as between the RIS and the mobile users. Structured covariance matrix interpolation optimally utilizes the array aperture from the sparsely placed active RIS elements. Recognizing the need for frequent channel estimation, we introduce a recurrent neural network-based model for sequential channel prediction, resulting in a significant reduction of the required training pilot signals. Simulation results affirm the capability and effectiveness of the proposed approach to enhance data transmission.

关键词： Channel estimation reconfigurable intelligent surface sparse array neural network sequential learning

来源：评论

学校读者我要写书评

暂无评论

BBSN: Bilateral-Branch Siamese network for Imbalanced Multi-label Text Classification 29th

BBSN: Bilateral-Branch Siamese Network for Imbalanced Multi-...

引用

29th International Conference on neural Information processing

作者： Zhao, Jiangjiang Li, Jiyz Fukumoto, Fumiyo Hangzhou Dianzi Univ Hangzhou Peoples R China Univ Yamanashi Kofu Yamanashi Japan

ISBN: (纸本)9783031301100;9783031301117

In multi-label text classification, the numbers of instances in different categories are usually extremely imbalanced. How to learn good models from imbalanced data is a challenging task. Some existing works tackle it through class re-balancing strategies or imbalanced loss objectives, but their performance remains limited in the cases of imbalanced distributed data. In this work, we propose a model, which combined Siamese network and Bilateral-Branch network to deal with both representation learning and classifier learning simultaneously. In the siamese network component, we propose a category-specific similarity strategy to improve the representation learning and adapt a novelty dynamic learning mechanism to make the model end-to-end trainable, and in the bilateral-branch network, we adopt the cumulative learning strategy to shift the learning focus from universal pattern to tail learning. In general, we adopt a multi-task architecture to ensure that both the head categories and the tail categories are adequately trained. The experiments on two benchmark datasets show that our method can improve the performance on the entire and tail categories, and achieves competitive performance compared with existing approaches.

关键词： Multi-label Text Classification Imbalanced Data

来源：评论

学校读者我要写书评

暂无评论

Mineral deposit grade assessment using a hybrid model of kriging and generalized regression neural network

引用

neural COMPUTING & APPLICATIONS 2022年第13期34卷 10611-10627页

作者： Singh, Rahul K. Ray, D. Sarkar, B. C. Indian Sch Mines Indian Inst Technol Dept Appl Geol Dhanbad 826004 Bihar India Indian Sch Mines Automat Ctr Indian Inst Technol Dhanbad 826004 Bihar India

Artificial neural networks are powerful global approximators for mineral grade assessment. The techniques are capable of retaining nonlinearity and spatial heterogeneity of a feature variable and can even perform modelling with noisy and incomplete data. These qualifiers make neural networks strong contenders for mineral grade assessment. In the present research, the authors have proposed a hybrid model consisting of two unsupervised models, namely ordinary kriging (OK) and k-means clustering, used as pre-processing steps to feed data to supervised generalized regression neural network (GRNN) model. While OK models the spatial variability and provides estimation on local scale, k-means clustering has been used for dimensionality reduction. These steps support in achieving accuracy and speed of the modelling process. GRNN model prepares its training and validation datasets using the k-clusters. Testing and validation of the model have been carried out on five live iron-ore deposits. Once the validation is found adequate and acceptable generalization with validation dataset is achieved, the model is verified with testing dataset. Deposit-wise value of R-2 of the hybrid GRNN model has been found to vary between 0.93 and 0.99. The model is observed to deliver improved performance when compared with multi-layer perceptron, radial basis function and recurrent neural network models. Spatially distributed estimation maps generated retain nonlinearity and spatial heterogeneity of the original Fe data values. Thus, the hybrid GRNN model provides an edge over the standalone classical geostatistics or standalone GRNN model.

关键词： Mineral grade assessment Ordinary kriging Generalized regression neural network Hybrid model

来源：评论

学校读者我要写书评

暂无评论

WH²D²N²: distributed AI-enabled OK-ASN Service forWeb of Things

引用

ACM TRANSACTIONS ON ASIAN AND LOW-RESOURCE LANGUAGE INFORMATION processing 2023年第5期22卷 1-16页

作者： Liang, Kun Ma, Ruhui Hua, Yang Wang, Hao Hu, Ningxin Song, Tao Gao, Honghao Guan, Haibing Shanghai Jiao Tong Univ Shanghai Peoples R China Queens Univ Belfas Belfast North Ireland Louisiana State Univ Baton Rouge LA USA Intel Corp Shanghai Minhang Off Shanghai Peoples R China Shanghai Univ Shanghai Peoples R China

Model data-driven ontology and knowledge presentation for evolving semantic Asian social networks (OKASN) is a critical strategy for web of things (WoT) services. Meanwhile, Deep neural network (DNN)-based OK-ASN service in WoT is growing rapidly. However, most DNN-based services cannot utilize the potential of WoT fully, as heterogeneity exists in WoT. Therefore, this article proposes a novel framework called Webbased HeterogeneousHierarchical distributedDeepneuralnetwork (WH2D2N2) to deploy the DNNs forOKASN services on WoT, overcoming the heterogeneity. The architecture of the system and the designed EdgeCloud-Joint execute scheme utilize heterogeneous devices to make DNN inference ubiquitous and output two types of results to meet various requirements. To bring robustness to OK-ASN services, a global scheduling is designed to arrange the workflow dynamically. The results of our experiments prove the efficiency of the execute scheme and the global scheduling in the system.

关键词： distributed inference web ecology heterogeneity ubiquitousness

来源：评论

学校读者我要写书评

暂无评论

Tokenization Representation and Deep-Learning-Based Intrusion Detection in Internet of Vehicles

引用

IEEE INTERNET OF THINGS JOURNAL 2024年第23期11卷 37974-37987页

作者： Gao, Jiaqi Lu, Yueming He, Yaru Fan, Mingrui Han, Daoqi Qiao, Yaojun Beijing Univ Posts & Telecommun Sch Informat & Commun Engn Beijing 100876 Peoples R China Beijing Univ Posts & Telecommun Key Lab Trustworthy Distributed Comp & Serv Minist Educ Beijing 100876 Peoples R China

The development of the Internet of Vehicles (IoV) has significantly enhanced connectivity and cooperation among road entities, leading to a more efficient, economical, and safer intelligent transportation system (ITS). However, this increased connectivity also exposes vehicles to a growing risk of cybersecurity threats through intravehicle and intervehicle networks. To secure IoV networks, many studies have focused on using intrusion detection systems (IDSs) based on deep learning methods to effectively detect cyber-attacks due to their ability to learn from large-scale data. Nonetheless, most existing IDSs rely on expert knowledge to manually design features, resulting in difficulties adapting to evolving attacks and information loss. To mitigate these limitations, this article presents a tokenization representation and attention mechanism-based convolutional neural network-bidirectional long short-term memory (CNN-BiLSTM) intrusion detection method. For feature extraction, we tokenize original traffic using natural language processing technique to represent discrete hexadecimal bytes as words, thus alleviating the need for manual feature design and allowing for the direct extraction of sequence patterns. For classification, we incorporate an attention mechanism into the CNN and BiLSTM architectures to enhance the accuracy of intrusion detection by focusing on critical information and capturing sequential patterns. The effectiveness of the proposed IDS is evaluated in both intravehicle and intervehicle network scenarios. Experimental results show that our method can detect various types of attacks with 100% accuracy on the Car-Hacking data set for the intravehicle network scenario. In the intervehicle network scenario utilizing the CICIoT2023 data set, our approach also achieves a high accuracy of 98%, outperforming existing methods.

关键词： intelligent transportation system (ITS) Internet of Things Internet of Vehicles (IoV) Deep learning intrusion detection intrusion detection traffic analysis traffic analysis Internet of Vehicles (IoV) intrusion detection traffic analysis

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：