检索结果-内蒙古大学图书馆

IEEE International Conference on Big Data and Cloud Computing (BdCloud)

作者： Wei Jiang Rui Xu Sheng Ma Qiong Wang Xiang Hou Hongyi Lu Science and Technology on Parallel and Distributed Processing Laboratory College of Computer National University of Defense Technology Changsha China

The correctness and robustness of the neural network model are usually proportional to its depth and width. Currently, the neural network models become deeper and wider to cope with complex applications, which leads to high memory capacity requirement and computer capacity requirements of the training process. The multi-accelerator parallelism is a promising choice for the two challenges, which deploys multiple accelerators in parallel for training neural networks. Among them, the pipeline parallel scheme has a great advantage in training speed, but its memory capacity requirements are relatively higher than other parallel schemes. Aiming at solving this challenge of pipeline parallel scheme, we propose a data transfer mechanism, which effectively reduces the peak memory usage of the training process by real-time data transferring. In the experiment, we implement our design and apply it to Pipedream, a mature pipeline parallel scheme. The memory requirement of training process is reduced by up to 48.5%, and the speed loss is kept within a reasonable range.

关键词： Training Computational modeling Pipelines Neural networks Memory management Bandwidth parallel processing

来源：评论

学校读者我要写书评

暂无评论

A modified estimation scheme of EBT algorithm for RFID system

A modified estimation scheme of EBT algorithm for RFID syste...

引用

International Conference on Intelligent Control and Information processing (ICICIP)

作者： Ruisi Zhao Yulu Yang Parallel and distributed processing laboratory Department of Computer Science and Technology Nankai University Tianjin China

The paper proposes a scheme to improve the accuracy of tag estimation in the EBT algorithm in RFID system. In the proposed scheme, the mean value of the estimated tag number should be acquired before calculating the optimal prefix. And the circular queue is used to record the average tag number which has been acquired recently. Thought the computer simulation we proposed the optimal length of the queue and make an explanation why the circular queue should be used to store the average tag number. The proposed scheme makes a better system performance compared with the original EBT algorithm.

关键词： Estimation Heuristic algorithms Algorithm design and analysis Radiofrequency identification Binary trees Accuracy Computer simulation

来源：评论

学校读者我要写书评

暂无评论

Adaptive simulations of viscous and inviscid detonations in supersonic combustible mixture 52nd

Adaptive simulations of viscous and inviscid detonations in ...

引用

52nd AIAA/SAE/ASEE Joint Propulsion Conference, 2016

作者： Cai, Xiaodong Liang, Jianhan Deiterding, Ralf Che, Yonggang Science and Technology on Scramjet Laboratory National University of Defense Technology ChangshaHunan410073 China Aerodynamics and Flight Mechanics Research Group University of Southampton Highfield Campus SouthamptonSO17 1BJ United Kingdom Science and Technology on Parallel and Distributed Processing Laboratory National University of Defense Technology ChangshaHunan410073 China

ISBN: (纸本)9781624104060

Two-dimensional high-resolution inviscid and viscous detonations were conducted in the supersonic combustible mixture with the open-source program AMROC. The results show that as the grid resolution increases, more small-scale vortices along the slip lines are generated in the inviscid detonation compared with the viscous one. The physical viscosity plays a major role in the whole diffusion effect, indicating that with the increasing of the grid resolution the viscous detonation can become convergent more easily. The diffusion effect brought by the physical viscosity can enhance the mixing of the unburned mixture with the combusted product, hence accelerating the consumption of the unburned reactant and the chemical heat release, and thus finally generating a stronger detonation than the inviscid one. The transverse wave of the viscous detonation can propagate faster, resulting in a lower periodic of the triple point oscillation than that of the inviscid one, mainly owing to the faster heat release resulting from the stronger diffusion effect. It is suggested that the strong diffusion effect involved in the viscous detonation can make a more significant impact on the suppression of the instability in the detonation front, thus eventually generating less triple points than the inviscid one, even with a larger overdrive degree. © 2016, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.

关键词： Detonation

来源：评论

学校读者我要写书评

暂无评论

Partial Clones for Stragglers in MapReduce

Partial Clones for Stragglers in MapReduce

引用

International Conference of Young Computer Scientists, Engineers and Educators, ICYCSEE 2015

作者： Jia Li Changjian Wang Dongsheng Li Zhen Huang National Laboratory for Parallel and Distributed Processing School of Computer ScienceNational University of Defense Technology

Stragglers can temporize jobs and reduce cluster efficiency *** researches have been contributed to the solution,such as Blacklist[8],speculative execution[1,6],Dolly[8].In this paper,we put forward a new approach for mitigating stragglers in Map Reduce,name *** starts task clones only for high-risk delaying *** experiments have been carried and results show that it can decrease the job delaying risk with fewer resources *** small jobs,Hummer also improves job completion time by 48% and 10% compared to LATE and Dolly.

关键词： MapReduce mitigating stragglers task clones

来源：评论

学校读者我要写书评

暂无评论

Co-designing the Topology/Algorithm to Accelerate Distributed Training

Co-designing the Topology/Algorithm to Accelerate Distribute...

引用

IEEE International Conference on Big Data and Cloud Computing (BdCloud)

作者： Xiang Hou Rui Xu Sheng Ma Qiong Wang Wei Jiang Hongyi Lu Science and Technology on Parallel and Distributed Processing Laboratory College of Computer National University of Defense Technology Changsha China

With the development of Deep Learning (DL), Deep Neural Network (DNN) models have become more complex. At the same time, the development of the Internet makes it easy to obtain large data sets for DL training. Large-scale model parameters and training data enhance the level of AI by improving the accuracy of DNN models. But on the other hand, they also present more severe challenges to the hardware training platform because training a large model needs a lot of computing and memory resources that can easily exceed the capacity of a single processor. In this context, integrating more processors on a hierarchical system to conduct distributed training is a direction for the development of training platforms. In distributed training, collective communication operations (including all-to-all, all-reduce, and all-gather) take up a lot of training time, making the interconnection network between computing nodes one of the most critical factors affecting the system performance. The hierarchical torus topology, combined with the Ring All-Reduce collective communication algorithm, is one of the current mainstream distributed interconnection networks. However, we believe that its communication performance is not the best. In this work, we first designed a new intra-package communication topology, i.e. the switch-based fully connected topology, which shortens the time consumed by cross-node communication. Then, considering the characteristics of this topology, we carefully devised more efficient all-reduce and all-gather communication algorithms. Finally, combined with the torus topology, we implemented a novel distributed DL training platform. Compared with the hierarchical torus, our platform improves communication efficiency and provides 1.16-2.68 times speedup in distributed training of DNN models.

关键词： Training Deep learning Network topology Multiprocessor interconnection Computational modeling Training data Switches

来源：评论

学校读者我要写书评

暂无评论

Outlier detection ensemble with embedded feature delection

arXiv

引用

arXiv 2020年

作者： Cheng, Li Wang, Yijie Liu, Xinwang Li, Bin Science and Technology on Parallel and Distributed Processing Laboratory College of Computer National University of Defense Technology Changsha China

Feature selection places an important role in improving the performance of outlier detection, especially for noisy data. Existing methods usually perform feature selection and outlier scoring separately, which would select feature subsets that may not optimally serve for outlier detection, leading to unsatisfying performance. In this paper, we propose an outlier detection ensemble framework with embedded feature selection (ODEFS), to address this issue. Specifically, for each random sub-sampling based learning component, ODEFS unifies feature selection and outlier detection into a pairwise ranking formulation to learn feature subsets that are tailored for the outlier detection method. Moreover, we adopt the thresholded self-paced learning to simultaneously optimize feature selection and example selection, which is helpful to improve the reliability of the training set. After that, we design an alternate algorithm with proved convergence to solve the resultant optimization problem. In addition, we analyze the generalization error bound of the proposed framework, which provides theoretical guarantee on the method and insightful practical guidance. Comprehensive experimental results on 12 real-world datasets from diverse domains validate the superiority of the proposed ODEFS. Copyright © 2020, The Authors. All rights reserved.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Smoothing Point Adjustment-Based Evaluation of Time Series Anomaly Detection

Smoothing Point Adjustment-Based Evaluation of Time Series A...

引用

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

作者： Mingyu Liu Yijie Wang Hongzuo Xu Xiaohui Zhou Bin Li Yongjun Wang Science and Technology on Parallel and Distributed Processing Laboratory College of Computer National University of Defense Technology Changsha China

Anomalies in time series appear consecutively, forming anomaly segments. Applying the classical point-based evaluation metrics to evaluate the detection performance of segments leads to considerable underestimation, so most related studies resort to point adjustment. This operation treats all points as true positives within a segment equally when only one individual point alarms, resulting in significant overestimation and creating an illusion of superior performance. This paper proposes smoothing point adjustment, a novel range-based evaluation protocol for time series anomaly detection. Our protocol reflects detection performance impartially by carefully considering the specific location and frequency of alarms in the raw results. It is achieved by smoothly determining the adjustment range and rewarding early detection via a ranging function and a rewarding function. Compared with other evaluation metrics, experiments on different datasets show that our protocol can yield a performance ranking of various methods more consistent with the desired situation.

关键词： Protocols Smoothing methods Time series analysis Detectors Signal processing Distance measurement Acoustics

来源：评论

学校读者我要写书评

暂无评论

Anode-ALM: An Aggregation-Based Application Level Multicast for Media Streaming

Anode-ALM: An Aggregation-Based Application Level Multicast ...

引用

International Symposium on parallel Architectures, Algorithms and Networks (ISPAN)

作者： Feng Liu Yuxing Peng Xicheng Lu Parallel and Distributed Processing Laboratory School of Computer Science and Technology National University of Defense Technology Changsha China

Building application level multicast (ALM) for media streaming, we had better keep the invariability of source to-end delay (S2EDelay) of any node on the tree, so as to avoid delay jitter problem. In this paper, we proposed an aggregation-based ALM architecture, which aggregates the unstable nodes within a local area into a relative stable aggregation node Anode, and then builds tree by Anodes. Anode shields node's dynamicity inside and behaves as a stable one outside. To build the architecture, we first proposed an Anode structure by n-dimension anti-cube, and proposed a distributed algorithm to bind the logic view and physical view of an Anode dynamically by the differentiation and assimilation operations, which maintain the integrality of the logic view while the physical view changes dynamically. Experiment results indicate that our algorithm can keep the S2EDelay invariability of the built ALM tree effectively.

关键词： Streaming media Anodes Delay Stability Application software Jitter Distributed algorithms Logic Internet Network topology

来源：评论

学校读者我要写书评

暂无评论

Word Embedding-based Context-sensitive Network Flow Payload Anomaly Detection

Word Embedding-based Context-sensitive Network Flow Payload ...

引用

International Conference on Applied Machine Learning (ICAML)

作者： Yizhou Li Yijie Wang Li Cheng Hongzuo Xu Science and Technology on Parallel and Distributed Processing Laboratory College of Computer National University of Defense Technology Changsha China

ISBN: (纸本)9781665421263

Payload anomaly detection can discover malicious behaviors hidden in network packets. It is hard to handle payload due to its various possible characters and complex semantic context, and thus identifying abnormal payload is also a non-trivial task. Prior art only uses the n-gram language model to extract features, which directly leads to ultra-high-dimensional feature space and also fails to capture the context semantics fully. Accordingly, this paper proposes a word embedding-based context-sensitive network flow payload anomaly detection method (termed WECAD). First, WECAD obtains the initial feature representation of the payload through the word embedding-based method. Then, we propose a corpus pruning algorithm, which applies the cosine similarity clustering and frequency distribution to prune inconsequential characters. We only keep the essential characters to reduce the calculation space. Subsequently, we propose a context learning algorithm. It employs the co-occurrence matrix transformation technology and introduces the backward step size to consider the order relationship of essential characters. Comprehensive experiments on real-world intrusion detection datasets validate the effectiveness of our method.

关键词： Art Semantics Clustering algorithms Intrusion detection Machine learning Feature extraction Task analysis

来源：评论

学校读者我要写书评

暂无评论

Surrogate Supervision-based Deep Weakly-supervised Anomaly Detection

Surrogate Supervision-based Deep Weakly-supervised Anomaly D...

引用

IEEE International Conference on Data Mining Workshops (ICDM Workshops)

作者： Zhiyue Wu Hongzuo Xu Yijie Wang Yongjun Wang Science and Technology on Parallel and Distributed Processing Laboratory College of Computer National University of Defense Technology Changsha China

ISBN: (纸本)9781665424288

Many anomaly detection applications can provide partially observed anomalies, but only limited work is for this setting. Additionally, a number of anomaly detectors focus on learning a particular model of normal/abnormal class. However, the intra-class model might be too complicated to be accurately learned. It is still a non-trivial task to handle data with anomalies/inliers in skewed and heterogeneous distributions. To address these problems, this paper proposes an anomaly detection method to leverage Partially Labeled anomalies via Surrogate supervision-based Deviation learning (denominated PLSD). The original supervision (i.e., known anomalies and a set of explored inliers) is transferred to semantic-rich surrogate supervision signals (i.e., anomaly-inlier and inlier-inlier class) via vector concatenation. Then different relationships and interactions between anomalies and inliers are directly and efficiently learned thanks to the neural network’s connection property. Anomaly scoring is processed via the trained network and the high-efficacy inliers. Extensive experiments show that PLSD significantly prevails state-of-the-art semi/weakly-supervised anomaly detectors.

关键词： Learning systems Conferences Detectors Gaussian distribution Benchmark testing Task analysis Anomaly detection

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：