检索结果-内蒙古大学图书馆

2024 Annual Modeling and Simulation Conference, ANNSIM 2024

作者： López, Jorge Abhari, Abdolreza Toronto Metropolitan University Canada

ISBN: (纸本)9781713899310

Adjusting hyperparameters, splitting data in deep neural network models, and adding more nodes to speed up are time-consuming tasks for real-world systems. We employ a novel Deep neural network (DNN) simulator to evaluate the effectiveness of parallelization via data splitting. While recent studies demonstrate successful DNN parallelization with abundant training data, there is a gap in simulators exploring proposed parallelization methods. This paper presents the DSMP simulator, a novel distributed system simulator that uses DNNs to assess training efficiency and accuracy through parallel nodes. Our simulated parallelization model distributes data across n-nodes, harnessing the power of distributed computing to improve model robustness and accuracy. By performing 10-fold Cross-Validation (10 fCV) in a distributed manner, our approach allows for the simultaneous processing of multiple folds, enhancing training efficiency. © 2024 Society for Modeling & Simulation International (SCS).

关键词： Digital elevation model

来源：评论

学校读者我要写书评

暂无评论

Response Time Analysis and Optimal Priority Assignment for Global Non-Preemptive Fixed-Priority Rigid Gang Scheduling

引用

IEEE TRANSACTIONS ON PARALLEL AND distributed SYSTEMS 2025年第3期36卷 455-470页

作者： Sun, Binqi Kloda, Tomasz Chen, Jiyang Lu, Cen Caccamo, Marco Tech Univ Munich D-85748 Munich Germany Univ Toulouse AAS CNRS INSA F-31000 Toulouse France

Non-preemptive rigid gang scheduling combines the efficiency of parallel execution with the reduced overhead of non-preemptive scheduling. This approach is particularly advantageous for parallel hardware accelerators, such as Google's Edge Tensor processing Unit (TPU), which is widely used for deep neural network (DNN) inference on embedded systems. This paper studies sporadic global non-preemptive fixed-priority (NP-FP) rigid gang scheduling, which is well-suited for DNN applications in Edge TPU pipelines. Each gang task spawns a fixed number of threads that must execute concurrently across distinct processing units. We introduce the first carry-in limitation technique specifically designed for gang task response time analysis, addressing the unique challenges posed by intra-task parallelism. This technique is formulated as a generalized knapsack problem, and we develop both a linear programming relaxation and a dynamic programming approach to solve it under different time complexities. Additionally, we propose the first optimal priority assignment policy for NP-FP gang schedulability tests. Our proposed schedulability analysis and optimal priority assignment policy are evaluated through extensive experiments, including both synthetic task sets and a case study using DNN benchmarks on commercial off-the-shelf Edge TPU accelerators. The results demonstrate that the proposed approaches effectively enhance the state-of-the-art global NP-FP gang schedulability tests, achieving improvements of up to 57.9% for synthetic task sets and 76.7% for Edge TPU benchmarks. Furthermore, we conduct an ablations study to examine the impact of different algorithmic components in the proposed technique, providing valuable insights for future research.

关键词： Optimal scheduling Artificial neural networks Scheduling Time factors Pipeline processing Scheduling algorithms Real-time systems Interference Benchmark testing Sun Gang scheduling response time analysis optimal priority assignment Edge TPU real-time system

来源：评论

学校读者我要写书评

暂无评论

Fast Policy Convergence for Traffic Engineering with Proactive distributed Message-Passing 38

Fast Policy Convergence for Traffic Engineering with Proacti...

引用

International Parallel and distributed processing Symposium (IPDPS)

作者： Wang, Zicheng Zhuang, Zirui Wang, Jingyu Qi, Qi Sun, Haifeng Liao, Jianxin Beijing Univ Posts & Telecommun Beijing Peoples R China

ISBN: (纸本)9798350387117;9798350387124

Nowadays, the rise of various network applications makes network traffic become increasingly complex, which brings more stringent requirements to traffic engineering (TE). Although the state-of-the-art TE approaches based on deep reinforcement learning (DRL) or traditional methods can generate optimal solutions for fixed traffic matrices, they cannot converge fast enough to provide real-time optimization in real networks either because of excessive computation times or high communication overheads. Moreover, due to the dynamically changing traffic load on the network, it is also challenging to achieve optimization of maximum link utilization (MLU) and end-to-end delay at the same time since these two optimization objectives may be conflicting, especially when the network is under a low traffic load, which makes the modeling very difficult. To meet these challenges, we present RT-TE, a TE system based on DRL and distributed message-passing between intelligent agents that can achieve real-time optimization for both MLU and end-to-end delay. To reduce the communication time due to link propagation delay during the optimization process, we design a proactive message-passing mechanism that allows agents to use partial messages to compute the routing policy while maintaining the optimization performance. Additionally, to achieve the tradeoff between the two optimization objectives, we model the propagation delay into the DRL model and design a multi-objective training framework with parameter transfer for training. Based on theoretical modeling, we can find the best tradeoff between the two objectives. Moreover, to improve the model's generalization for various traffic flows, we use a GNN model to generate the rewards of the DRL model, which greatly speeds up the training phase and allows us to feed massive amounts of data into the model. Through evaluations of real-world network topologies, our approach shows a 10%-20% improvement in optimizing MLU under short traffic-ch

关键词： traffic engineering distributed networked system deep reinforcement learning message passing graph neural network

来源：评论

学校读者我要写书评

暂无评论

distributed Inference in Resource-Constrained IoT for Real-Time Video Surveillance

引用

IEEE SYSTEMS JOURNAL 2023年第1期17卷 1512-1523页

作者： Khan, Muhammad Asif Hamila, Ridha Erbad, Aiman Gabbouj, Moncef Qatar Univ Doha Qatar Qatar Univ Dept Elect Engn Doha Qatar Hamad Bin Khalifa Univ Doha Qatar Tampere Univ Tampere 33100 Finland

Advances in communication technologies and computational capabilities of Internet of Things (IoT) devices enable a range of complex applications that require ever increasing processing of sensors' data. An illustrative example is real-time video surveillance that captures videos of target scenes and process them to detect anomalies using deep learning (DL). Running deep learning models requires huge processing and incurs high computation delay and energy consumption on resource-constraint IoT devices. In this article, we introduce methods for distributed inference over IoT devices and edge server. Two distinct algorithms are proposed to split the deep neural network layers computation between IoT device and an edge server;the early split strategy (ESS) for battery powered IoT devices and the late split strategy (LSS) for IoT devices connected to regular power source. The evaluation shows that both the ESS and LSS schemes achieve the target inference delay deadline when tested over VGG16 and MobileNet_V2 CNN models. In terms of computational load, the ESS scheme achieves nearly 15-20% reduction whereas LSS scheme achieves up to 60% reduction. The gains in energy saving of IoT devices for both the ESS and LSS schemes are nearly 18% and 52%, respectively.

关键词： Computational modeling Internet of Things Servers Task analysis Energy consumption Delays Load modeling Computation distributed machine learning edge computing inference neural networks split strategy video IoT

来源：评论

学校读者我要写书评

暂无评论

Ballastless track arching recognition based on one-dimensional residual convolutional neural network and vehicle response

引用

CONSTRUCTION AND BUILDING MATERIALS 2023年 408卷

作者： Tang, Xueyang Chen, Zelin Cai, Xiaopei Wang, Yi Beijing Jiaotong Univ Sch Civil Engn Beijing 100044 Peoples R China

In the long-term service process, high-temperature load will lead to ballastless track arching. The ballastless track arching will reduce the structural integrity, affect the smoothness of the line, and even endanger the safety of driving. Existing detection methods are costly, complex and not suitable for large-scale. This study proposes a low-cost, simple detection method based on one-dimensional residual convolutional neural network and vehicle response to recognize the arching. The vehicle responses in the simulation model are used as inputs to compare the recognition accuracies of different algorithms for the arching, and the vehicle response features extracted by the optimal algorithm are visualized in 3D using t-distributed Stochastic Neighbor Embedding. The results show that the vehicle vertical acceleration is sensitive to the arching amplitude, and the influence of the arching on the vehicle vertical acceleration ranges from about 40 m, and, the vehicle vertical acceleration in the range of 40 m is chosen as the input. The network 2-2 containing two residual blocks 2 has the highest recognition accuracy. The accuracy at different vehicle speeds ranged from 96.5 % to 97.5 %, and the average accuracy at a single arching magnitude ranged from 95.0 % to 100 %. The original vehicle response features under different arching mag-nitudes are very complexly distributed in 3D space with more overlapping. After network 2-2 processing, the vehicle response features corresponding to different arching magnitudes are basically separated in 3D space, and only very few features overlap. It shows that network 2-2 has strong robustness and vehicle response feature extraction capability.

关键词： High -speed railway 1DCNN Residual network Vehicle response Ballastless track arching

来源：评论

学校读者我要写书评

暂无评论

Exploiting Sparsity in Pruned neural networks to Optimize Large Model Training 37

Exploiting Sparsity in Pruned Neural Networks to Optimize La...

引用

37th IEEE International Parallel and distributed processing Symposium (IPDPS)

作者： Singh, Siddharth Bhatele, Abhinav Univ Maryland Dept Comp Sci College Pk MD 20742 USA

ISBN: (纸本)9798350337662

Parallel training of neural networks at scale is challenging due to significant overheads arising from communication. Recently, deep learning researchers have developed a variety of pruning algorithms that are capable of pruning (i.e. setting to zero) 80-90% of the parameters in a neural network to yield sparse subnetworks that equal the accuracy of the unpruned parent network. In this work, we propose a novel approach that exploits these sparse subnetworks to optimize the memory utilization and communication in two popular algorithms for parallel deep learning namely - data and inter-layer parallelism. We integrate our approach into AxoNN, a highly scalable framework for parallel deep learning that relies on data and inter-layer parallelism, and demonstrate the reduction in communication time and memory utilization. On 512 NVIDIA V100 GPUs, our optimizations reduce the memory consumption of a 2.7 billion parameter model by 74%, and the total communication time by 40%, thus providing an overall speedup of 34% over AxoNN, 32% over DeepSpeed-3D and 46% over Sputnik, a sparse matrix computation baseline.

关键词： lottery ticket hypothesis sparse computations GPUs parallel deep learning memory optimizations

来源：评论

学校读者我要写书评

暂无评论

Incremental learning of upper limb action pattern recognition based on mechanomyography

引用

BIOMEDICAL SIGNAL processing AND CONTROL 2023年第Part1期79卷 000-000页

作者： Zhao, Tongtong Cao, Gangsheng Zhang, Yue Zhang, Hanyang Xia, Chunming East China Univ Sci & Technol Sch Mech & Power Engn Shanghai 200237 Peoples R China Nantong Univ Sch Mech Engn Jiangsu Prov 226001 China Nantong Peoples R China Shanghai Univ Engn Sci Sch Mech & Automot Engn Shanghai 201620 Peoples R China

Aiming at the problem that the traditional mechanomyography (MMG) pattern recognition model for upper limb rehabilitation action has poor recognition effect on non-identical distribution test data, this research proposes an incremental learning method for deep learning based on MMG. By collecting 8-channels of MMG of 4 types of hand movements (wrist flexion, wrist extension, wrist ulnar flexion, and wrist radial flexion), after signal pre-processing, feature extraction and dimensionality reduction, 12 groups of non-identical distributed data were obtained. The model was trained by using deep neural network (DNN), and five commonly used machine learning algorithms were used as comparison for incremental training. Finally, the recognition rate of DNN was 88.25%, and the final recognition rates of Passive Aggressive (PA), Incremental Support Vector Machine (ISVM), Perceptron, Bernoulli Naive Bayes (BNB) and Multinomial Naive Bayes (MNB) were 85.94%, 84.82%, 81.01%, 73.07% and 61.20% respectively. Among them, both DNN and PA had better upward trends, and DNN had the highest final recognition rate. By comparing the confusion matrices before and after DNN incremental learning, it could be seen that DNN incremental training can significantly improve the accuracy and precision in confusion matrix. The experimental results demonstrate that the method of incremental learning can not only achieve the recognition of non-identically distributed test data, but also improve the recognition rates and the generalization performance of the model.

关键词： Upper limb rehabilitation action Mechanomyography Non -identical distribution Incremental learning Deep neural network

来源：评论

学校读者我要写书评

暂无评论

Distribution prediction of moisture content of dead fuel on the forest floor of Maoershan national forest, China using a LoRa wireless network

引用

Journal of Forestry Research 2022年第3期33卷 899-909页

作者： Bo Peng Jiawei Zhang Jian Xing Jiuqing Liu Northeast Forestry University Harbin150040People’s Republic of China

The moisture content of dead forest fuel is an important indicator of risk levels of forest fires and prediction of fire spread. Moisture distribution is important to determine wild fire rating. However, it is often difficult to predict moisture distribution because of a complex terrain, changeable environments and low cover of commercial communication signals inside the forest. This study proposes a moisture content prediction system composed of environmental data collected using a long range radio frequency band 433 MHz wireless sensor network and data processing for moisture prediction based on a BP (back-propagation) neural network. In the fall of 2019, twenty nodes for the collection of environmental data were placed in four forest stands of Maoershan National Forest for a month;7440 sets of data including temperature, humidity, wind speed and air pressure were obtained. Half the data were used as a training set, the other as a testing set for a BP neural network. The results show that the average absolute error between the predicted value and the real value of moisture content of fuels of Larix gmelini, Betula platyphylla, Juglans mandshurica, and Quercus mongolica stands was 0.94%, 0.21%, 0.86%, 0.97%, respectively. The prediction accuracy was relatively high. The proposed distributed moisture content prediction method has the advantages of wide coverage and good real-time performance;at the same time, it is not limited by commercial signals and so it is especially suitable for forest fire prediction in remote mountainous areas.

关键词： distributed moisture content prediction Dead fuel BP neural network

来源：评论

学校读者我要写书评

暂无评论

Federated Learning for network Traffic Prediction 23

Federated Learning for Network Traffic Prediction

引用

23rd International-Federation-for-Information-processing (IFIP) networking Conference (IFIP networking)

作者： Behera, Sadananda Panda, Saroj Kumar Panayiotou, Tania Ellinas, Georgios Natl Inst Technol Rourkela Dept Elect & Commun Engn Sundargarh Odisha India LTIMindtree Bangalore India Univ Cyprus KIOS CoE Nicosia Cyprus Univ Cyprus Dept Elect & Comp Engn Nicosia Cyprus

ISBN: (纸本)9798350390605;9783903176638

network traffic prediction is imperative for effective network planning, decision-making, and optimization, leveraging the inherent predictability observed in traffic patterns. Although various machine learning models, such as deep neural networks (NNs) with recurrent units, including long-short-term memory (LSTM) NNs, have shown promise in this area, they are conventionally centralized trained, overlooking the potential benefits of distributed training techniques at the network edge. Exploiting advances in distributed training could provide certain advantages to the problem at hand, such as improved privacy preservation, reduced processing times, and lower bandwidth usage. Thus, this work proposes a federated learning (FL) framework for predicting network traffic, utilizing LSTMs for local model training. Simulation results validate the effectiveness of the proposed model, as this distributed implementation demonstrates high traffic prediction accuracy on real traffic traces, comparable to the traditional centralized approach, while safeguarding data privacy.

关键词： Federated learning network Traffic Prediction Machine Learning

来源：评论

学校读者我要写书评

暂无评论

Collaborative Prediction in Anti-Fraud System Over Multiple Credit Loan Platforms

引用

IEEE TRANSACTIONS ON DEPENDABLE AND SECURE COMPUTING 2024年第4期21卷 3580-3596页

作者： Wang, Cheng Tang, Hao Zhu, Hangyu Jiang, Changjun Tongji Univ Dept Comp Sci & Technol Shanghai 201804 Peoples R China Minist Educ Key Lab Embedded Syst & Serv Comp Shanghai 201804 Peoples R China Shanghai Artificial Intelligence Lab Shanghai 200030 Peoples R China

Anti-fraud engineering for online credit loan (OCL) platforms is getting more challenging due to the developing specialization of gang fraud. Associations are critical features referring to assessing the credibility of loan applications for OCL fraud prediction. State-of-the-art solutions employ graph-based methods to mine hidden associations among loan applications effectively. They perform well based on the information asymmetry which is guaranteed by the huge advantage of platforms over fraudsters in terms of data quantity and quality at their disposal. The inherent difficulty that can be foreseen is the data isolation caused by mistrust between multiple platforms and data control legislations for privacy preservation. To maintain the advantage owned by the platforms, we design a privacy-preserving distributed graph learning framework that ensures critical association repairs by merging parameter sharing and data sharing. Specially, we propose the association reconstruction mechanism (ARM) that consists of the devised exploration, processing, transmission and utilization schemes to realize data sharing. For parameter sharing, we design a hybrid encryption technique to protect privacy during collaboratively learning graph neural network (GNN) models among different financial client platforms. We conduct the experiments over real-life data from large financial platforms. The results demonstrate the effectiveness and efficiency of our proposed methods.

关键词： Fraud Business Servers Privacy distributed databases Data privacy Graph neural networks Fraud prediction online credit loan distributed graph learning federated learning

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：