检索结果-内蒙古大学图书馆

21st International Conference on Harmonics and Quality of Power

作者： You, Bo Xing, Cheng Yu, Qiao Zhang, Maolin He, Zhaoran Zeng, Lizhen Liu, Chen Datang Chongqing Energy Mkt CO LTD Chongqing Peoples R China Chongqing Univ Sch Elect Engn Chongqing Peoples R China

ISBN: (纸本)9798350382570;9798350382563

With the increasing penetration of distributed energy resources at the grid edge, including renewable energy generation, flexible loads, and energy storage devices, accurately predicting consumer-level distributed generation and consumption has become crucial. However, conventional centralized processing and machine learning approaches are impractical because of data security and privacy issues. This paper proposes a load forecasting method based on federated learning with LSTM neural network which updates model parameters by having clients responsible for model training and a server responsible for model aggregation, allowing multiple clients to collaboratively train a machine learning model without sharing raw data. Simulation results show that this method can achieve accurate predictions while preserving privacy.

关键词： Data privacy distributed energy resources federated Learning load forecasting LSTM neural network

来源：评论

学校读者我要写书评

暂无评论

Swift : Expedited Failure Recovery for Large-Scale DNN Training

引用

IEEE TRANSACTIONS ON PARALLEL AND distributed SYSTEMS 2024年第9期35卷 1644-1656页

作者： Zhong, Yuchen Sheng, Guangming Liu, Juncheng Yuan, Jinhui Wu, Chuan Univ Hong Kong Hong Kong Peoples R China OneFlow Inc Beijing 100089 Peoples R China

As the size of deep learning models gets larger and larger, training takes longer time and more resources, making fault tolerance more and more critical. Existing state-of-the-art methods like CheckFreq and Elastic Horovod need to back up a copy of the model state (i.e., parameters and optimizer states) in memory, which is costly for large models and leads to non-trivial overhead. This article presents Swift, a novel recovery design for distributed deep neural network training that significantly reduces the failure recovery overhead without affecting training throughput and model accuracy. Instead of making an additional copy of the model state, Swift resolves the inconsistencies of the model state caused by the failure and exploits the replicas of the model state in data parallelism for failure recovery. We propose a logging-based approach when replicas are unavailable, which records intermediate data and replays the computation to recover the lost state upon a failure. The re-computation is distributed across multiple machines to accelerate failure recovery further. We also log intermediate data selectively, exploring the trade-off between recovery time and intermediate data storage overhead. Evaluations show that Swift significantly reduces the failure recovery time and achieves similar or better training throughput during failure-free execution compared to state-of-the-art methods without degrading final model accuracy. Swift can also achieve up to 1.16x speedup in total training time compared to state-of-the-art methods.

关键词： Training Parallel processing Pipelines Computational modeling Data models Artificial neural networks Checkpointing distributed DNN training failure resilience

来源：评论

学校读者我要写书评

暂无评论

Synchronization of Coupled Memristive neural network Based on Edge-Event Triggered Control

引用

neural processing LETTERS 2023年第8期55卷 11209-11232页

作者： An, Letian Yang, Yongqing Hao, Rixu Li, Li Jiangnan Univ Sch Sci Xue Lang St Wuxi 214122 Jiangsu Peoples R China Jiangnan Univ Sch Internet Things Xue Lang St Wuxi 214122 Jiangsu Peoples R China

This research aims to investigate synchronization issues in coupled memristive neural networks (CMNNs) using both the static and dynamic edge-event triggered control protocols. An interval parameter system is developed by integrating the concept of Filippov solution with differential inclusion theory. Unlike existing work, the suggested edge-event triggered mechanisms don't require the constant information transfer among neighboring nodes, providing a more distributed control approach that reduces system resources since each node communicates asynchronously. Additionally the absence of Zeno behavior at any given moment supports the efficacy of the approach. To demonstrate its viability, a practical simulation example is presented.

关键词： Edge-event triggered control protocol Synchronization Coupled memristive neural networks Zeno behavior

来源：评论

学校读者我要写书评

暂无评论

A Multidimensional Communication Scheduling Method for Hybrid Parallel DNN Training

引用

IEEE TRANSACTIONS ON PARALLEL AND distributed SYSTEMS 2024年第8期35卷 1415-1428页

作者： Li, Shengwei Lu, Kai Lai, Zhiquan Liu, Weijie Ge, Keshi Li, Dongsheng Natl Univ Def Technol Coll Comp Natl Key Lab Parallel & Distributed Comp Changsha 410073 Peoples R China

The transformer-based deep neural network (DNN) models have shown considerable success across diverse tasks, prompting widespread adoption of distributed training methods such as data parallelism and pipeline parallelism. With the increasing parameter number, hybrid parallel training becomes imperative to scale training. The primary bottleneck in scaling remains the communication overhead. The communication scheduling technique, emphasizing the overlap of communication with computation, has demonstrated its benefits in scaling. However, most existing works focus on data parallelism, overlooking the nuances of hybrid parallel training. In this paper, we propose TriRace, an efficient communication scheduling framework for accelerating communications in hybrid parallel training of asynchronous pipeline parallelism and data parallelism. To achieve effective computation-communication overlap, TriRace introduces 3D communication scheduling, which adeptly leverages data dependencies between communication and computations, efficiently scheduling AllReduce communication, sparse communication, and peer-to-peer communication in hybrid parallel training. To avoid possible communication contentions, TriRace also incorporates a topology-aware runtime which optimizes the execution of communication operations by considering ongoing communication operations and real-time network status. We have implemented a prototype of TriRace based on PyTorch and Pipedream-2BW, and conducted comprehensive evaluations with three representative baselines. Experimental results show that TriRace achieves up to 1.07-1.45x speedup compared to the state-of-the-art pipeline parallelism training baseline Pipedream-2BW, and 1.24-1.81x speedup compared to the Megatron.

关键词： Training Computational modeling Processor scheduling Pipelines Data models Transformers Pipeline processing distributed training deep learning hybrid parallelism communication scheduling

来源：评论

学校读者我要写书评

暂无评论

RCIF: TOWARDS ROBUST distributed DNN COLLABORATIVE INFERENCE UNDER HIGHLY LOSSY networkS 49

RCIF: TOWARDS ROBUST DISTRIBUTED DNN COLLABORATIVE INFERENCE...

引用

49th IEEE International Conference on Acoustics, Speech, and Signal processing (ICASSP)

作者： Cheng, Yujun Zhang, Zhewei Wang, Shengjin Tsinghua Univ Dept Elect Engn Beijing Peoples R China Beijing Natl Res Ctr Informat Sci & Technol BNRis Beijing Peoples R China ZEEWAIN Technol Co Ltd Beijing Peoples R China

ISBN: (纸本)9798350344868;9798350344851

Collaborative Inference is a prospective paradigm for accelerating Deep neural network (DNN) inference by harnessing the computational resources of multiple devices. However, in highly lossy network environments, such as those encountered in wireless communication systems, the transmission loss of intermediate feature maps between devices can result in significant degradation of co-inference accuracy. In this paper, we first conduct a comprehensive investigation into the impact of intermediate feature map loss in real-world wireless scenarios and provide an in-depth analysis of loss patterns under UDP transmission. Motivated by these observations, we introduce Robust Co-inference Framework (RCIF), a novel framework that employs a hierarchical mask strategy to selectively drop activations at two different scales of feature maps. This approach enhances the robustness of DNN co-inference in the presence of network losses. Our evaluation on a variety of datasets and network architectures demonstrates that RCIF significantly enhances the accuracy and robustness of distributed DNN co-inference under highly lossy network conditions. Specifically, our results show that RCIF can achieve up to a 659% increase in accuracy compared to the original model under particularly poor network conditions.

关键词： Collaborative Inference neural networks Lossy Transmission Deep Learning Robustness

来源：评论

学校读者我要写书评

暂无评论

Trustworthy distributed Deep neural network Training in an Edge Device network

Trustworthy Distributed Deep Neural Network Training in an E...

引用

IEEE International Conference on Big Data

作者： Sudipta Saha Shubha Haiying Shen Department of Computer Science University of Virginia USA

ISBN: (纸本)9781665480468

With the increased usage of edge devices having local computation capabilities, deep neural network (DNN) training in a network of edge devices becomes promising. Several recent works have proposed fully edge-based distributed training systems for situations when the communication to cloud is unstable or intermittent. However, such distributed systems become vulnerable when there are untrusted devices that launch data and model poisoning attacks during training, deteriorating the accuracy of the DNN model. To handle this challenge, we propose a Trustworthy distributed system for Machine learning training in an edge device network (TrustMe). TrustMe realizes both data and model parallelisms. It detects the untrusted devices producing illegitimate outputs. Next, it reassigns the training tasks of the untrusted devices to other trusted devices in such a way that the reassignment and the training that is restarted after the reassignment require minimal time. Our container-based emulation and real device experiments demonstrate that TrustMe achieves up to 12% higher accuracy and 45% less training time compared to existing methods in the presence of untrusted devices.

关键词： Training Deep learning Image edge detection neural networks Emulation distributed databases Parallel processing

来源：评论

学校读者我要写书评

暂无评论

A Hybrid Model Based on CNN-LSTM to Detect and Forecast Harmonics: A Case Study of an Eskom Substation in South Africa

引用

ELECTRIC POWER COMPONENTS AND SYSTEMS 2023年第8期51卷 746-760页

作者： Kuyumani, E. M. Hasan, Ali N. Shongwe, T. Univ Johannesburg Dept Elect & Elect Engn Sci Johannesburg South Africa

The ever-growing modern smart grid with more distributed energy resources is providing efficient energy supply while facing several challenges that include harmonics induced among many. Previous and present literature shows that various machine and deep learning models are superior and accurate as compared to the traditional and conventional signal processing techniques. Obtaining accurate results becomes extremely important especially the fact that harmonics are essentially nonlinear, nonparametric, and adaptive in nature. This paper proposes a novel forecasting model that aggregates two deep learning models: convolutional neural network (CNN) and long short term memory (LSTM) recurrent neural network (RNN) detect and forecast harmonics in a power system. CNN-LSTM hybrid forecasting model for harmonics in the power grid system has achieved significantly superior performance in collaborative data mining on spatiotemporal measurement data. Sample features are extracted using CNN before they are passed through LSTM for prediction. To show the superiority of the hybrid CNN-LSTM deep neural prediction network model, it is compared with CNN, LSTM and NARX (Non-Linear Autoregressive with External (Exogenous) Input). CNN-LSTM forecasting performance is superior as compared to the other four models. MSE and RMSE for CNN-LSTM are 0.00038 ([3.8 x 10] omicron (-4)) and 0.0000014917 (1.4917 x 10 omicron(-6)) respectively.

关键词： harmonics convolutional neural network (CNN) long short term memory (LSTM) non-linear autoregressive with external (exogenous) input (NARX) MSE RMSE

来源：评论

学校读者我要写书评

暂无评论

Auditing of Hadoop log file for dynamic detection of threats using H-ISSM-MIM and Convolutional neural network

引用

JOURNAL OF INTELLIGENT & FUZZY SYSTEMS 2023年第4期45卷 6617-6628页

作者： Suganya, S. Selvamuthukumaran, S. SASTRA Deemed Univ Dept CSE Kumbakonam Tamil Nadu India AVC Coll Engn Dept Comp Applicat Mayiladuthurai Tamil Nadu India

Hadoop is a big data processing system that enables the distributed processing of massive data sets across multiple computers using straightforward programming techniques. Hadoop has been extensively investigated in many attacks as a result of its growing significance in industry. A company may learn about the actions of invaders as well as the weaknesses of the Hadoop cluster by examining a significant quantity of data from the log file. In a Big Data setting, the goal of the paper is to generate an analytical classification for intrusion detection. In this study, Hadoop log files were examined based on assaults that were recorded in the log files. Prior to analysis, the log data is cleaned and improved using a Hadoop preprocessing tool. For feature extraction, the hybrid Improved Sparrow Search Algorithm with Mutual Information Maximization (H-ISSA-MIM). Then the CNN (Convolutional neural network) classifier will detect the intrusions. The implementation is performed using the MATLAB 2020a software. The performance metrics like accuracy, precision, F-score, recall, specificity, FPR, FNR are calculated for the proposed methodology and it is compared with the existing techniques like Decision Tree (DT), Principal Components Analysis (PCA)-K means, Long Short Time Memory (LSTM). The maximum value of accuracy finds out in the proposed method 98%.

关键词： Hadoop attacks log file intrusion detection big data environment and feature extraction convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Feature Extraction and Classification Techniques for Power Quality Disturbances in distributed Generation: A Review

引用

IETE JOURNAL OF RESEARCH 2023年第6期69卷 3836-3851页

作者： Singh, Nivedita Ansari, M. A. Tripathy, Manoj Singh, Vivek Pratap Gautam Buddha Univ Dept Elect Engn Greater Noida 201312 India IIT Roorkee Dept Elect Engn Roorkee 247667 Uttar Pradesh India

The purpose of this paper is to analyze the frequent power quality (PQ) issues happening in distributed generation, the outcomes of the PQ harmonics, the methods used to assess the quantity of harmonic distortion which occurs in the power system (PS), and, in the end, classification of these disturbances using recent advance artificial intelligent techniques like a neural network, fuzzy logic, and the genetic algorithm has been further stated. To protect the PS detection and classification of voltage (V) and current (I) issues are essential tasks and due to increasing interest in a distributed generation, it is becoming more popular. Most PQ disturbances are unstable and ephemeral especially in a distributed generation;therefore, the call for detection and classification of voltage and current disruptions are essential tasks to protect the PS. Many disturbances of (PQ are unpredictable and transient. By using wavelet transforms, expert systems, and artificial neural networks, some intelligent system technologies control fault analysis precisely saying it can help to detect the fault locations. The most important part of the generalized classification system of PQ events is the extraction and classification of features for PQ event classification.

关键词： Classifiers distributed generation (DG) intelligent techniques power quality (PQ) power quality improvement device (PQI) random forest (RF) signal processing techniques

来源：评论

学校读者我要写书评

暂无评论

Bandwidth-efficient distributed neural network architectures with application to body sensor networks

arXiv

引用

arXiv 2022年

作者： Strypsteen, Thomas Bertrand, Alexander STADIUS Center for Dynamical Systems Signal Processing and Data Analytics Leuven.AI - KU Leuven institute for AI Kasteelpark Arenberg 10 LeuvenB-3001 Belgium

In this paper, we describe a conceptual design methodology to design distributed neural network architectures that can perform efficient inference within sensor networks with communication bandwidth constraints. The different sensor channels are distributed across multiple sensor devices, which have to exchange data over bandwidth-limited communication channels to solve, e.g., a classification task. Our design methodology starts from a user-defined centralized neural network and transforms it into a distributed architecture in which the channels are distributed over different nodes. The distributed network consists of two parallel branches of which the outputs are fused at the fusion center. The first branch collects classification results from local, node-specific classifiers while the second branch compresses each node’s signal and then reconstructs the multichannel time series for classification at the fusion center. We further improve bandwidth gains by dynamically activating the compression path when the local classifications do not suffice. We validate this method on a motor execution task in an emulated EEG sensor network and analyze the resulting bandwidth-accuracy trade-offs. Our experiments show that the proposed framework enables up to a factor 20 in bandwidth reduction with minimal loss (up to 2%) in classification accuracy compared to the centralized baseline on the demonstrated motor execution task. The proposed method offers a way to smoothly transform a centralized architecture to a distributed, bandwidth-efficient network amenable for low-power sensor networks. While the application focus of this paper is on wearable brain-computer interfaces, the proposed methodology can be applied in other sensor network-like applications as well. Copyright © 2022, The Authors. 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

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：