首页 | 官方网站   微博 | 高级检索  
相似文献
 共查询到20条相似文献,搜索用时 15 毫秒
1.
Java垃圾收集机制及性能调节   总被引:2,自引:0,他引:2  
Java技术通过自动的垃圾收集来管理内存的回收,降低了编写代码的复杂程度,减小了系统中潜在的不安全因素。针对众多开发人员在实际应用中未能充分发挥垃圾收集机制优势的情况,基于对Java虚拟机中垃圾收集机制及其实现的分析,提出了在具体应用中对垃圾收集算法进行选择和对Java虚拟机相关参数进行调节的方法。实验表明,该方法使得垃圾收集对系统性能的负面影响最小化,从而提高了系统的性能和效率。对开发人员具有一定的参考价值。  相似文献   

2.
3.
Nowadays, clustered environments are commonly used in high‐performance computing and enterprise‐level applications to achieve faster response time and higher throughput than single machine environments. Nevertheless, how to effectively manage the workloads in these clusters has become a new challenge. As a load balancer is typically used to distribute the workload among the cluster's nodes, multiple research efforts have concentrated on enhancing the capabilities of load balancers. Our previous work presented a novel adaptive load balancing strategy (TRINI) that improves the performance of a clustered Java system by avoiding the performance impacts of major garbage collection, which is an important cause of performance degradation in Java. The aim of this paper is to strengthen the validation of TRINI by extending its experimental evaluation in terms of generality, scalability and reliability. Our results have shown that TRINI can achieve significant performance improvements, as well as a consistent behaviour, when it is applied to a set of commonly used load balancing algorithms, demonstrating its generality. TRINI also proved to be scalable across different cluster sizes, as its performance improvements did not noticeably degrade when increasing the cluster size. Finally, TRINI exhibited reliable behaviour over extended time periods, introducing only a small overhead to the cluster in such conditions. These results offer practitioners a valuable reference regarding the benefits that a load balancing strategy, based on garbage collection, can bring to a clustered Java system. Copyright © 2016 John Wiley & Sons, Ltd.  相似文献   

4.
张聪品  吴长茂  赵理莉 《计算机应用》2010,30(11):2876-2879
为了提高垃圾收集效率,减少用户程序等待时间,提出了一种在多核系统下基于LISP2算法的并行节点复制算法。该算法通过把LISP2算法的4个垃圾收集阶段分别并行化来实现并行垃圾收集。实验结果显示,该算法在多核系统下能有效提高垃圾收集效率。  相似文献   

5.
This paper investigates the time requirements for an implementation of retention block-structured languages that uses a garbage collector as its sole means for recovering inaccessible storage. The usual three-pass markcompactify-and-update garbage collector is optimized to eliminate the need for the third pass in the event that the executed program is lifetime well-stacking.Supported (in part) by the United States Energy Research and Development Administration, Contract No. E(04-3)-34, PA 214, and (in part) by the National Science Foundation, Grant No. DCR75-08659.  相似文献   

6.
针对城市生活垃圾分类收运过程中存在的环境二次污染和垃圾产生量不确定性等问题,提出了一种基于智能垃圾桶的动态收运车辆路径优化方法。建立以最小化碳排放成本、燃油消耗成本、固定成本和车辆延迟到达惩罚成本为目标的动态车辆路径优化模型。采用滚动时域的方式将动态问题转换为一系列静态问题,并设计两阶段算法进行求解。首先采用粒子群算法对收运车辆路径进行规划,而后在每个时域末,综合考虑待清运垃圾桶的位置和垃圾量、垃圾收运车辆的位置和装载量以动态调整现有车辆路径。研究结果表明,相较于传统的静态收运方案,动态垃圾收运方案能够在降低车辆运输成本和碳排放成本的同时,显著降低由于清运不及时造成的环境二次污染的风险。  相似文献   

7.
非增量式Java虚拟机(JVM)垃圾回收算法的内存开销较大。为此,提出一种基于栈式分配策略的JVM增量式垃圾收集算法。对Java栈帧进行改造使其支持存储对象,改进增量式收集器中堆空间的划分、引用跟踪方式,以减少垃圾收集带来的不确定性暂停。实验结果表明,该算法能有效减少暂停的频率和时长,提高运行速度。  相似文献   

8.
王粲  夏元清  邹伟东 《计算机应用研究》2021,38(6):1724-1727,1764
针对极限学习机(extreme learning machine,ELM)隐节点不确定性导致的系统不稳定,以及对大型数据计算负担过重的问题,提出了基于自适应动量优化算法(adaptive and momentum method,AdaMom)的正则化极限学习机.算法主要思想是构造连续可微的目标函数,在梯度下降过程中计算自适应学习率,求自适应学习率与梯度乘积的指数加权平均值,通过迭代得到损失函数最小值对应的隐层输出权重矩阵.实验结果表明,在相同基准数据集的训练中,AdaMom-ELM算法具有非常良好的泛化性能和鲁棒性,提高了计算效率.  相似文献   

9.
江海洋 《计算机应用研究》2010,27(12):4430-4432
提出了一种新的方法挖掘评论中的文字信息,将评论对象被用户关注的层面发掘出来并评分,根据这些层面的分数以及用户过往的评分数据学习出用户的偏好,最后根据用户的偏好预测其他待评分对象的分数并产生推荐。实验结果表明,提出的方法在预测准确度方面较传统方法有一定程度的提高。  相似文献   

10.
11.
To maximize visualization effectiveness, graphical data are commonly augmented with text to provide detailed information and define specific values. This text is often displayed in a pop-up dialog box pertaining to an object, permitting simultaneous display of the object and associated alphanumeric information. However, a human operator performance cost may be incurred when the resulting portrayal occludes critical information within the visualization. To address this issue, we developed and evaluated three alternative, spatially constrained, text portrayal techniques. These techniques and their associated access interface were designed to reduce occlusion while providing rapid access to desired alphanumeric data. Each technique was evaluated against the pop-up dialog using a dual-task human performance paradigm. Performance measures included accuracy, response time, display occlusion, and subjective feedback. The basis for the spatially-constrained text access technique design, their implementation affordances, and limitations are discussed. The alternative techniques and their user interface concepts resulted in mixed accuracy and response time performance compared to the pop-up dialog. Specific design features reduced data access time to one third of the time required to access the spatially-constrained text access techniques. Overall, equivalent performance was obtained among the variants while the potential for occlusion was reduced during use of the novel designs.  相似文献   

12.

基于大数据机器学习的智能软件研发过程需要综合运用软件工程、数据与领域知识工程、机器学习等多方面的知识和工具,涉及的研究主题和人员角色众多,技术实现手段复杂、研发难度大. 面向智能软件的需求工程需要面对领域知识、业务知识、数据科学交织带来的挑战. 然而,如何将领域知识和端到端的机器学习技术恰当地融合到给定的业务流程之中,以及如何应对工业、医疗等高可信要求场景中的可解释性需求,仍是亟待探索的重要研究问题. 调研了近年来面向机器学习应用的需求工程研究文献,对该领域的发展现状、核心问题和代表性方法进行综述. 提出了面向机器学习应用的可解释性需求分析框架. 基于该框架,通过一个工业智能应用案例分析了未来待研究的重要问题,展望了可行的研究路径.

  相似文献   

13.
Abstract. A general way of contrasting the object-oriented to the function-oriented approach, called 'object-order' and 'function-order', is proposed in this paper. In addition, when modelling information systems, there is a third principle of ordering, 'subject-order'. Object and function-order are used to define the two approaches to modelling information systems and in human-computer interaction, in ways that are compatible to the corresponding distinction in programming. The conditions for selecting either of the two approaches are, however, not discussed here.
By means of an object-ordered method, e.g. Jackson System Development (JSD), an object-ordered model is produced, and the user interface that can be derived from the model is object-ordered too. However, other guidelines in JSD support subject-order. A function-ordered interface can be derived from the subject-ordered model. A computer system designed by means of an object-organized method may therefore obtain an interface that is partially object-ordered and partially function-ordered. Other object-oriented methods give guidelines for designing function-oriented interfaces without relating these guidelines to the model.
The system designer may benefit from having three principles of ordering when modelling a system. However, a user may be confused by an interface that sometimes gives preference to objects, sometimes to functions.  相似文献   

14.
提出了一种评价云中心综合性能指标的分析架构.基于性能代理和服务接口方式(PASI)建立分析架构,该接口方式由性能客户端(PMC)、性能代理(PMA)和性能服务端(PMS)组成(GAS),并提出了一个基于排队原理的数学模式来论证这个方法的可行性.实验结果表明,以PASI模式对云中心性能指标进行采集和评估的方式是非常有效的.  相似文献   

15.
由于用户在实际使用云平台时,很难确定云平台的云主机类型,所以造成了云平台资源利用率低下的问题。许多典型的解决资源利用率低下的方法,都是从云提供商的角度优化放置算法,而用户选择将限制资源利用率增加;也有一些方法采用云平台下的任务性能短时间采集并预测,但会降低任务分类的准确性。为了达到提高云平台资源利用率、简化用户操作的目的,首先提出一种多属性的任务性能采集工具Lbenchmark,全面采集任务的性能特征,和Ganglia相比负载降低了50%以上。然后,利用该性能数据,提出一种基于权值可配的多KD树-K最近邻(KNN)应用性能分类算法,挑选适合参数建立多个基于KD树的KNN分类器,通过交叉验证方法调整每个属性在不同分类器的权重,进行选举分类。实验结果表明,所提算法与传统的KNN相比,计算量明显提高了约10倍以上,而准确性平均提高约10%。该算法可利用数据特征映射将资源建议提供给用户和云提供商,进而提高云平台整体的利用率。  相似文献   

16.
A highly important part of software engineering education is requirements collection and analysis which is one of the initial stages of the Database Application Lifecycle and arguably the most important stage of the Software Development Lifecycle. No other conceptual work is as difficult to rectify at a later stage or as damaging to the overall system if performed incorrectly. As software engineering is a field with a reputation for producing graduates who are inappropriately prepared for applying their skills in real life software engineering scenarios, it suggests that traditional educational techniques such as role-play, live-through case studies and paper-based case studies are insufficient preparation and that other approaches are required. To attempt to combat this problem we have developed a games-based learning application to teach requirements collection and analysis at tertiary education level as games-based learning is seen as a highly motivating, engaging form of media and is a rapidly expanding field. This paper will describe the evaluation of the requirements collection and analysis game particularly from a pedagogical perspective. The game will be compared to traditional methods of software engineering education using a pre-test/post-test, control group/experimental group design to assess if the game can act as a suitable supplement to traditional techniques and assess if it can potentially overcome shortcomings. The game will be evaluated in five separate experiments at tertiary education level.  相似文献   

17.
李洪亮  张弄  孙婷  李想 《计算机应用》2022,42(6):1649-1655
通过分析分布式机器学习中作业性能干扰的问题,发现性能干扰是由于内存过载、带宽竞争等GPU资源分配不均导致的,为此设计并实现了快速预测作业间性能干扰的机制,该预测机制能够根据给定的GPU参数和作业类型自适应地预测作业干扰程度。首先,通过实验获取分布式机器学习作业运行时的GPU参数和干扰率,并分析出各类参数对性能干扰的影响;其次,依托多种预测技术建立GPU参数-干扰率模型进行作业干扰率误差分析;最后,建立自适应的作业干扰率预测算法,面向给定的设备环境和作业集合自动选择误差最小的预测模型,快速、准确地预测作业干扰率。选取5种常用的神经网络作业,在两种GPU设备上设计实验并进行结果分析。结果显示,所提出的自适应干扰预测(AIP)机制能够在不提供任何预先假设信息的前提下快速完成预测模型的选择和性能干扰预测,耗时在300 s以内,预测干扰率误差在2%~13%,可应用于作业调度和负载均衡等场景。  相似文献   

18.
极限学习机(ELM)在训练过程中无需调整隐层节点参数,因其高效的训练方式被广泛应用于分类和回归,然而极限学习机也面临着结构选择与过拟合等严重等问题。为了解决此问题,针对隐层节点增量数目对收敛速度以及训练时间的影响进行了研究,提出一种利用网络输出误差的变化率控制网络增长速度的变长增量型极限学习机算法(VI-ELM)。通过对多个数据集进行回归和分类问题分析实验,结果表明,本文提出的方法能够以更高效的训练方式获得良好的泛化性能。  相似文献   

19.
This paper presents a methodology to apply speech technology for compensating sensory, motor, cognitive and affective usage difficulties. It distinguishes (1) an analysis of accessibility and technological issues for the identification of context-dependent user needs and corresponding opportunities to include speech in multimodal user interfaces, and (2) an iterative generate-and-test process to refine the interface prototype and its design rationale. Best practices show that such inclusion of speech technology, although still imperfect in itself, can enhance both the functional and affective information and communication technology-experiences of specific user groups, such as persons with reading difficulties, hearing-impaired, intellectually disabled, children and older adults.
Khiet P. TruongEmail:
  相似文献   

20.
An increasingly widespread interest in developing fully adaptable e-learning systems (e.g., intelligent tutoring systems) has led to the development of a wide range of adaptive processes and techniques. In particular, advances in these systems are based on optimization for each user's learning style and characteristics, to enable a personalized learning experience. Current techniques are aimed at using a learner's personality traits and its effect on learning preferences to improve both the initial learning experience and the information retained (e.g., top-down or bottom-up learning organization). This study empirically tested the relationship between a learner's personality traits, analyzed the effects of these traits on learning preferences, and suggested design guidelines for adaptive learning systems. Two controlled experiments were carried out in a computer-based learning session. Our first experiment showed a significant difference in the learning performance of participants who were identified as introverts vs. those who were identified as being extroverts, according to the MBTI scale. As the distinction between extroverted personality types vs. introverted personality types showed the strongest correlation in terms of different learning styles, we used this criteria in our second experiment to determine whether design guidelines for appropriate content organization could reinforce the aforementioned correlation between personality type and learning experience.  相似文献   

设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号