检索结果-内蒙古大学图书馆

2018 ieee/acm Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems, PMBS 2018

作者： Malakar, Preeti Balaprakash, Prasanna Vishwanath, Venkatram Morozov, Vitali Kumaran, Kalyan Argonne National Laboratory United States

ISBN: (纸本)9781728101828

Performance modeling is an important and active area of research in high-performance computing (hpc). It helps in better job scheduling and also improves overall performance of coupled applications. Sufficiently rich analytical models are challenging to develop, however, because of interactions between different node components, network topologies, job interference, and application complexity. When analytical performance models become restrictive because of application dynamics and/or multicomponent interactions, machine-learning-based performance models can be helpful. While machine learning (ML) methods do not require underlying system or application knowledge, they are efficient in learning the unknown interactions of the application and system parameters empirically using application runs. We present a benchmark study in which we evaluate eleven machine learning methods for modeling the performance of four representative scientific applications that are irregular and with skewed domain configurations on four leadership-class hpc platforms. We assess the impact of feature engineering, size of training set, modern hardware platforms, transfer learning, extrapolation on the prediction accuracy, and training and inference times. We find that bagging, boosting, and deep neural network ML methods are promising approaches with median R2 values greater than 0.95 and these methods do not require feature engineering. We demonstrate that cross-platform performance prediction can be improved significantly using transfer learning with deep neural networks. © 2018 ieee.

关键词： Benchmarking

来源：评论

学校读者我要写书评

暂无评论

MPCA SGD-A Method for Distributed Training of Deep learning Models on Spark

引用

ieee TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS 2018年第11期29卷 2540-2556页

作者： Langer, Matthias Hall, Ashley He, Zhen Rahayu, Wenny La Trobe Univ Melbourne Vic 3086 Australia La Trobe Univ Dept Comp Sci Melbourne Vic 3086 Australia La Trobe Univ Comp Sci Melbourne Vic 3086 Australia La Trobe Univ Sch Engn & Math Sci Melbourne Vic 3086 Australia

Many distributed deep learning systems have been published over the past few years, often accompanied by impressive performance claims. In practice these figures are often achieved in high performance computing (hpc) environments with fast InfiniBand network connections. For average deep learning practitioners this is usually an unrealistic scenario, since they cannot afford access to these facilities. Simple re-implementations of algorithms such as EASGD [1] for standard Ethernet environments often fail to replicate the scalability and performance of the original works [2]. In this paper, we explore this particular problem domain and present MPCA SGD, a method for distributed training of deep neural networks that is specifically designed to run in low-budget environments. MPCA SGD tries to make the best possible use of available resources, and can operate well if network bandwidth is constrained. Furthermore, MPCA SGD runs on top of the popular Apache Spark [3] framework. Thus, it can easily be deployed in existing data centers and office environments where Spark is already used. When training large deep learning models in a gigabit Ethernet cluster, MPCA SGD achieves significantly faster convergence rates than many popular alternatives. For example, MPCA SGD can train ResNet-152 [4] up to 5.3x faster than state-of-the-art systems like MXNet [5], up to 5.3x faster than bulk-synchronous systems like SparkNet [6] and up to 5.3x faster than decentral asynchronous systems like EASGD [1].

关键词： Deep learning distributed computing machine learning neural networks spark stochastic gradient descent

来源：评论

学校读者我要写书评

暂无评论

A Novel Architecture for Automatic Document Classification for Effective Security in Edge Computing environments 3

A Novel Architecture for Automatic Document Classification f...

引用

3rd ieee/acm Symposium on Edge Computing (SEC)

作者： Ding, Lei Ben Salem, Malek Accenture Labs Arlington VA 22209 USA

ISBN: (纸本)9781538694459

Existing automatic document classification architectures are underpinned by the transfer of raw document data streams to the cloud computing environment for processing and analysis. This operation is expensive, has limitation to meet the real-time processing needs, and exposes potential security and privacy risks at the data transfer process. In this paper, we present a novel architecture for document classification with enhanced security that incorporates mechanism to facilitate the processing and analysis of raw data streams near the source of the data.

关键词： Edge Computing Security Document Classification machine learning Data Classification Edge Classification

来源：评论

学校读者我要写书评

暂无评论

A scalable framework for online power modelling of high-performance computing nodes in production 16

A scalable framework for online power modelling of high-perf...

引用

International Conference on High Performance Computing & Simulation (hpcS)

作者： Pittino, Federico Beneventi, Francesco Bartolini, Andrea Benini, Luca Univ Bologna Dept Elect Elect & Informat Engn DEI Bologna Italy Swiss Fed Inst Technol Integrated Syst Lab Zurich Switzerland

ISBN: (纸本)9781538678794

Power and thermal design and management are critical components of high performance computing (hpc) systems, due to their cutting-edge position in terms of high power density and large total power consumption. Many hpc power management strategies rely on the availability of accurate compact power models, capable of predicting power consumption and tracking its sensitivity to workload parameters and operating points. In this paper we describe a methodology and a framework for training power models derived with two of the best-in-class procedures directly on the online in production nodes and without requiring dedicated training instances. The compact power models are obtained using an online regression-based approach which can track non-stationary workloads and hardware variability. Our experiments on a real-life hpc system demonstrate that the models achieve very high accuracy over all operating modes. We also demonstrate the scalability of our approach and the small amount of resources needed for the online modeling, for both the training and inference phases.

关键词： power model hpc cluster in production machine learning scalable framework

来源：评论

学校读者我要写书评

暂无评论

Virtualized GPUs in High Performance Datacenters 16

Virtualized GPUs in High Performance Datacenters

引用

International Conference on High Performance Computing and Simulation (hpcS)

作者： Kurkure, Uday Vu, Lan Sivaraman, Hari VMware Palo Alto CA 94304 USA

ISBN: (纸本)9781538678794

Virtualization has become the foundation of high performance datacenters. The scale of high performance datacenters is similar to the scale of supercomputers used in High Performance Computing (hpc). Both involve millions of CPU cores, tens of thousands of accelerators like GPUs, FPGAs connected by high speed interconnects and different connection topologies. GPUs are essential in high performance computing because of their high throughput and energy efficiency. Many high performance computing workloads have avoided virtualization because of performance concerns. As the scale of high performance datacenter increases, the need for ease of management becomes critical. The virtualized GPUs offer new opportunities in efficient resource allocation in managing datacenters. In this paper, we present different GPU virtualization solutions and discuss its related aspects that helps GPU based applications running efficiently at scale. We also present a comprehensive performance study of virtualized GPU that addresses main considerations of running high performance applications in virtualized datacenters. We evaluate the performance of virtual GPUs from many different angles: the overhead of virtual GPUs compared to physical GPU, the impact of using container, the use of virtualized GPUs for many different types of hpc applications, the solution of mixing workloads to increase resource utilization and system consolidation, the scalability, as well as how to make the right choice of vGPU profile.

关键词： GPU High Performance Computing vGPU machine learning deep learning cloud computing performance optimization Virtualization Virtual machine Graphics

来源：评论

学校读者我要写书评

暂无评论

Online learning applied to Autonomous Valuation of Financial Assets 18

Online learning applied to Autonomous Valuation of Financial...

引用

ieee/WIC/acm International Conference on Web Intelligence (WI)

作者： Ribeiro, Marcel Soares Lima de Castro, Paulo Andre Technol Inst Aeronaut ITA Autonomous Computat Syst Lab Sao Jose Dos Campos SP Brazil

ISBN: (纸本)9781538673256

In the context of Artificial Intelligence, Online learning is focused on environments that are not independent and identically distributed, i.e. the environment may changes its behavior as time goes by. Blum proposed a famous algorithm to this problem, which was called randomized weighted majority algorithm. In this paper, we propose an adaptation of such algorithm to autonomous valuation of financial assets. Our approach is based on learning from expert's advices, in order to create a more adaptable solution and reuse some results achieved for other researchers. We also briefly review some papers in the field. The proposed approach is materialized through an online learning algorithm that defines an analysis derived from many different analyses performed by autonomous analysts. Such analysts may be created using techniques from finance or machine learning fields. Our algorithm is able to take into account different costs of analysis errors. We believe that this skill in fundamental to an efficient analyst. We implemented the algorithm and tested it using several different techniques from finance and one (very simple) algorithm from machine learning area. This implementation was tested and the achieved results are analyzed and discussed. Furthermore, we proved that our algorithm's cost of error is limited by an expression of the cost of error of the best autonomous analyst. We believe that this algorithm may contribute to development of better systems that intend to estimate the price of financial assets in an autonomous way.

关键词： Investment machine learning algorithms machine learning Probability distribution Computational modeling Analytical models Cost accounting

来源：评论

学校读者我要写书评

暂无评论

Poster Abstract: Comparison of Classifiers for Prediction of Human Actions in a Smart Home 3

Poster Abstract: Comparison of Classifiers for Prediction of...

引用

3rd acm/ieee International Conference on Internet of Things Design and Implementation (IoTDI)

作者： Alhafidh, Basman M. Hasan Daood, Amar, I Allen, William H. Florida Inst Technol Dept Elect & Comp Engn Melbourne FL 32901 USA Florida Inst Technol Sch Comp Melbourne FL 32901 USA

ISBN: (纸本)9781538663127

There is a strong interest in IoT-based systems that monitor and control smart home environments by accurately predicting the needs of the human occupants. Past research has focused on the accuracy of prediction of a user's future action. However, much of that work uses synthetic datasets which do not always reflect the real-world interactions that occur between an individual and the home environment. In addition, a focus on prediction accuracy often comes at the cost of slower processing time. This paper focuses on the prediction of future human actions in an intelligent environment with the goal of achieving both high accuracy and real-time performance. We performed experiments using the MavPad dataset, which was gathered from a fully-instrumented home environment, and compared several different machine learning algorithms that included both single and ensemble classifiers. The results show that using a Support Vector machine approach achieved the best results when using a group of sensors within a local zone around the user and the Random Forest classifier achieved higher performance when using sensors that are distributed across the entire home environment.

关键词： Smart Home System IoT Ensemble Classifier machine learning Algorithms Prediction System

来源：评论

学校读者我要写书评

暂无评论

Flux: Overcoming Scheduling Challenges for Exascale Workflows 13

Flux: Overcoming Scheduling Challenges for Exascale Workflow...

引用

13th ieee/acm Workshop on Workflows in Support of Large-Scale Science (WORKS)

作者： Ahn, Dong H. Bass, Ned Chu, Albert Garlick, Jim Grondona, Mark Herbein, Stephen Koning, Joseph Patki, Tapasya Scogland, Thomas R. W. Springmeyer, Becky Taufer, Michela Lawrence Livermore Natl Lab 7000 East Ave Livermore CA 94550 USA Univ Tennessee Knoxville TN USA

ISBN: (纸本)9781728101965

Many emerging scientific workflows that target high-end hpc systems require complex interplay with the resource and job management software (RJMS). However, portable, efficient and easy-to-use scheduling and execution of these workflows is still an unsolved problem. We present Flux, a novel, hierarchical RJMS infrastructure that addresses the key scheduling challenges of modern workflows in a scalable, easy-to-use, and portable manner. At the heart of Flux lies its ability to be nested seamlessly within batch allocations created by other schedulers as well as itself. Once a hierarchy of Flux instance is created within each allocation, its consistent and rich set of well-defined APIs portably and efficiently support those workflows that can often feature non-traditional execution patterns such as requirements for complex co-scheduling, massive ensembles of small jobs and coordination among jobs in an ensemble.

关键词： Computational modeling Cancer Data models machine learning Task analysis Software Resource management

来源：评论

学校读者我要写书评

暂无评论

Anatomy of a fog-to-cloud distributed recommendation system in airports 11

Anatomy of a fog-to-cloud distributed recommendation system ...

引用

11th ieee/acm International Conference on Utility and Cloud Computing (UCC-Companion) / 5th ieee/acm International Conference on Big Data Computing, Applications and Technologies (BDCAT)

作者： Salis, Antonio Bulla, Roberto Mancini, Glauco Cocco, Paolo Jensen, Jens Engn Sardegna Srl Telco & Media Div Cagliari Italy UKRI STFC Rutherford Appleton Lab Dept Comp Sci Chilton England

ISBN: (纸本)9781728103594

Airports are confusing environments even at the best of times. Recent advances in IoT have made it possible to develop real-time improved traveler assistance tools for mobile phones, which we describe in this paper. These tools will be particularly valuable for the elderly or disabled travelers, or people with other constraints that require assistance, but will be highly valuable also to the seasoned traveler. The tool covers the essential path through the airport onto the flight, but also allows the traveler to fill their waiting time by discovering the facilities of the airport, aided by a recommender system using machine-learning. At the same time the system provides the airport with anonymized user "heat maps" which can highlight congestion and other aggregated data to highlight situations that require intervention, such as emergencies. Initially targeting Android, the app makes use of the fog-to-cloud platform developed by the Horizon2020-funded mF2C project.

关键词： Cloud computing fog computing fog-to-cloud distributed systems IoT collaborative filtering machine learning proximity marketing

来源：评论

学校读者我要写书评

暂无评论

Ensemble learning for effective run-time hardware-based malware detection: A comprehensive analysis and classification 18

Ensemble learning for effective run-time hardware-based malw...

引用

55th Annual Design Automation Conference, DAC 2018

作者： Sayadi, Hossein Patel, Nisarg Sai Manoj, P.D. Sasan, Avesta Rafatirad, Setareh Homayoun, Houman George Mason University FairfaxVA22030 United States

ISBN: (纸本)9781450357005

Malware detection at the hardware level has emerged recently as a promising solution to improve the security of computing systems. Hardware-based malware detectors take advantage of machine learning (ML) classifiers to detect pattern of malicious applications at run-time. Šese ML classifiers are trained using low-level features such as processor Hardware Performance Counters (hpcs) data which are captured at run-time to appropriately represent the application behaviour. Recent studies show the potential of standard ML-based classifiers for detecting malware using analysis of large number of microarchitectural events, more than the very limited number of hpc registers available in todayfs microprocessors which varies from 2 to 8. Šis results in executing the application more than once to collect the required data, which in turn makes the solution less practical for effective run-time malware detection. Our results show a clear tradeoff between the performance of standard ML classifiers and the number and diversity of hpcs available in modern microprocessors. Šis paper proposes a machine learning-based solution to break this trade-off to realize effective run-time detection of malware. We propose ensemble learning techniques to improve the performance of the hardwarebased malware detectors despite using a very small number of microarchitectural events that are captured at run-time by existing hpcs, eliminating the need to run an application several times. For this purpose, eight robust machine learning models and two well-known ensemble learning classifiers applied on all studied ML models (sixteen in total) are implemented for malware detection and precisely compared and characterized in terms of detection accuracy, robustness, performance (accuracy robustness), and hardware overheads. Še experimental results show that the proposed ensemble learning-based malware detection with just 2 hpcs using ensemble technique outperforms standard classifiers with 8 hpcs by up to

关键词： Malware

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：