函数优化异步并行演化算法

提出了一种新型、高效的函数优化异步并行演化算法,利用这个算法,在巨型并行计算机上解决了一些高难度的大型优化问题,其中包括一个超高维的非线性规划问题－BUMP问题。由于BUMP问题的强非线性和超我峰特性,目前还未见有超过50维的BUMP问题的结果发表。而在此不仅仅得到了从2维到50维迄今最好的解,而且一直计算到了1000000维,并得到了满意的结果。数值实验表明,新算法是鲁棒和高效的。     

分布式并行粒子滤波算法结构分析与研究

粒子滤波器是解决非线性/非高斯系统状态估计的有效技术,广泛应用于目标跟踪、无线通信.自动控制等领域.但因其计算复杂、计算量庞大等缺陷,无法满足实时系统的应用需求.针对粒子滤波器计算量大、实时性差的问题,提出了一种基于MPI的分布式并行粒子滤波算法,给出了Master-Slave并行模式下任务分配、数据划分与负载平衡策略.实验结果表明,若忽略通信代价,加速比基本呈线性增长.     

图嵌入算法的分布式优化与实现

随着人工智能时代的到来,图嵌入技术被越来越多地用来挖掘图中的信息.然而,现实生活中的图通常很大,因此,分布式图嵌入技术得到了广泛的关注.分布式图嵌入算法面临着两大难点:(1)图嵌入算法多种多样,没有一个通用的框架能够描述大部分的算法;(2)现在的分布式图嵌入算法扩展性不足,当处理大图时性能较低.针对以上两个挑战,首先提...     

一种分布式并行参数优化算法设计及其应用

并行参数优化算法在科学计算中有广泛应用。随着Spark等分布式平台的快速发展,越来越多并行参数优化算法开始采用分布式平台进行实现。如何在Spark等平台上设计优化算法,避免其运行效率受到框架固定时间开销和网络I/O影响,已经成为亟需解决的问题。本文设计一种分布式与单机多核并行结合的参数优化算法,将其划分为调度部分和独立子问题部分,单机多核并行算法处理子问题,分布式平台负责子问题的跨节点并行。碳通量模型参数优化的实验结果表明,改进的算法能有效节省时间开销,更快地搜索参数空间。     

并行机器学习算法基础体系前沿进展综述

大数据环境下,机器学习算法受到前所未有的重视。总结和分析了传统机器学习算法在海量数据场景下出现的若干问题,基于当代并行机分类回顾了国内外并行机器学习算法的研究现状,并归纳总结了并行机器学习算法在各种基础体系下存在的问题。针对大数据环境下并行机器学习算法进行了简要的总结,并对其发展趋势作了展望。     

基于机器学习的网络故障诊断优化算法

随着网络的广泛应用,网络故障已成为影响用户体验和业务连续性的主要挑战之一。研究基于机器学习的网络故障诊断优化算法,旨在提高网络故障诊断的准确性和效率。该算法采用深度神经网络（Deep Neural Networks,DNN）模型,通过数据预处理、特征工程、网络故障诊断模型建立等步骤,并模拟真实网络环境进行实验验证,证明了其有效性和可行性。     

MUSIC算法在分布式并行处理机上的实现研究

介绍了一种基于分布式并行处理平台的并行MUSIC实现方法。在实现过程中,针对MUSIC算法运算量大难以实时实现的特点,采取了有效减少计算量的措施。并根据MUSIC算法各个子任务的不平衡性,提出了一种基于软件流水的并行任务划分方法。经仿真试验证明,经过上述并行处理后可有效加快算法的完成时间,在工程实践中具有广泛的实际意义。     

大数据下的机器学习算法综述

随着产业界数据量的爆炸式增长,大数据概念受到越来越多的关注.由于大数据的海量、复杂多样、变化快的特性,对于大数据环境下的应用问题,传统的在小数据上的机器学习算法很多已不再适用.因此,研究大数据环境下的机器学习算法成为学术界和产业界共同关注的话题.文中主要分析和总结当前用于处理大数据的机器学习算法的研究现状.此外,并行是处理大数据的主流方法,因此介绍一些并行算法,并引出大数据环境下机器学习研究所面临的问题.最后指出大数据机器学习的研究趋势.     

Boosting Algorithms for Parallel and Distributed Learning

The growing amount of available information and its distributed and heterogeneous nature has a major impact on the field of data mining. In this paper, we propose a framework for parallel and distributed boosting algorithms intended for efficient integrating specialized classifiers learned over very large, distributed and possibly heterogeneous databases that cannot fit into main computer memory. Boosting is a popular technique for constructing highly accurate classifier ensembles, where the classifiers are trained serially, with the weights on the training instances adaptively set according to the performance of previous classifiers. Our parallel boosting algorithm is designed for tightly coupled shared memory systems with a small number of processors, with an objective of achieving the maximal prediction accuracy in fewer iterations than boosting on a single processor. After all processors learn classifiers in parallel at each boosting round, they are combined according to the confidence of their prediction. Our distributed boosting algorithm is proposed primarily for learning from several disjoint data sites when the data cannot be merged together, although it can also be used for parallel learning where a massive data set is partitioned into several disjoint subsets for a more efficient analysis. At each boosting round, the proposed method combines classifiers from all sites and creates a classifier ensemble on each site. The final classifier is constructed as an ensemble of all classifier ensembles built on disjoint data sets. The new proposed methods applied to several data sets have shown that parallel boosting can achieve the same or even better prediction accuracy considerably faster than the standard sequential boosting. Results from the experiments also indicate that distributed boosting has comparable or slightly improved classification accuracy over standard boosting, while requiring much less memory and computational time since it uses smaller data sets.     

A Library Hierarchy for Implementing Scalable Parallel Search Algorithms

This paper describes the design of the Abstract Library for Parallel Search (ALPS), a framework for implementing scalable, parallel algorithms based on tree search. ALPS is specifically designed to support data-intensive algorithms, in which large amounts of data are required to describe each node in the search tree. Implementing such algorithms in a scalable manner is challenging both because of data storage requirements and communication overhead. ALPS incorporates a number of new ideas to address this challenge. The paper also describes the design of two other libraries forming a hierarchy built on top of ALPS. The first is the Branch, Constrain, and Price Software (BiCePS) library, a framework that supports the implementation of parallel branch and bound algorithms in which the bounds are obtained by solving some sort of relaxation, usually Lagrangian. In this layer, the notion of global data objects associated with the variables and constraints is introduced. These global objects provide a connection between the various subproblems in the search tree, but they pose further difficulties for designing scalable algorithms. The other library is the BiCePS linear integer solver (BLIS), a concretization of BiCePS, in which linear programming is used to obtain bounds in each search tree node.     

基于机器学习的智能路由算法综述

互联网的飞速发展催生了很多新型网络应用,其中包括实时多媒体流服务、远程云服务等.现有尽力而为的路由转发算法难以满足这些应用所带来的多样化的网络服务质量需求.随着近些年将机器学习方法应用于游戏、计算机视觉、自然语言处理获得了巨大的成功,很多人尝试基于机器学习方法去设计智能路由算法.相比于传统数学模型驱动的分布式路由算法而言,基于机器学习的路由算法通常是数据驱动的,这使得其能够适应动态变化的网络环境以及多样的性能评价指标优化需求.基于机器学习的数据驱动智能路由算法目前已经展示出了巨大的潜力,未来很有希望成为下一代互联网的重要组成部分.然而现有对于智能路由的研究仍然处于初步阶段.首先介绍了现有数据驱动智能路由算法的相关研究,展现了这些方法的核心思想和应用场景并分析了这些工作的优势与不足.分析表明,现有基于机器学习的智能路由算法研究主要针对算法原理,这些路由算法距离真实环境下部署仍然很遥远.因此接下来分析了不同的真实场景智能路由算法训练和部署方案并提出了2种合理的训练部署框架以使得智能路由算法能够低成本、高可靠性地在真实场景被部署.最后分析了基于机器学习的智能路由算法未来发展中所面临的机遇与挑战并给出了未来的研究方向.     

基于机器学习的动态分区并行文件系统性能优化

近年来,随着大数据、云计算技术的发展,应用系统越来越集中,规模亦越来越大,使得存 储系统的性能问题越来越突出。为应对其性能要求,并行文件系统得到了大量的应用。然而现有的并 行文件系统优化方法,大多只考虑应用系统或并行文件系统本身,较少考虑两者之间的协同。该文基 于应用系统在并行文件系统上的访问模式对存储系统的性能有显著影响这一特点,提出基于动态分区 的并行文件系统优化方法。首先,利用机器学习技术来分析挖掘各个性能影响因素和性能指标之间的 关系和规律,生成优化模型。其次,以优化模型为基础,辅助并行文件系统的参数调优工作。最后, 基于 Ceph 存储系统进行原型实现,并设计了三层架构应用系统进行了性能测试,最终达到优化并行 文件系统访问性能的目的。实验结果表明,所提出方法可以达到 85% 的预测优化准确率;在所提出模 型的辅助优化下,并行文件系统的吞吐量性能得到约 3.6 倍的提升。     

分布式存储环境下矩阵转置并行算法研究

本文针对基于消息传递机制的分布式高性能计算中经常遇到的矩阵转置问题,在分析环状算法的基础上,提出了一种新的蝶形并送算法。     

Scalable Parallel Matrix Multiplication on Distributed Memory Parallel Computers

Consider any known sequential algorithm for matrix multiplication over an arbitrary ring with time complexity O(N^α), where 2<α3. We show that such an algorithm can be parallelized on a distributed memory parallel computer (DMPC) in O(log N) time by using N^α/log N processors. Such a parallel computation is cost optimal and matches the performance of PRAM. Furthermore, our parallelization on a DMPC can be made fully scalable, that is, for all 1pN^α/log N, multiplying two N×N matrices can be performed by a DMPC with p processors in O(N^α/p) time, i.e., linear speedup and cost optimality can be achieved in the range [1..N^α/log N]. This unifies all known algorithms for matrix multiplication on DMPC, standard or non- standard, sequential or parallel. Extensions of our methods and results to other parallel systems are also presented. For instance, for all 1p N^α /log N, multiplying two N×N matrices can be performed by p processors connected by a hypercubic network in O(N^α/p+(N²/p^2/α)(log p)^2(α−1)/α) time, which implies that if p=O(N^α/(log N)^{2(α−1)/(α−2)}), linear speedup can be achieved. Such a parallelization is highly scalable. The above claims result in significant progress in scalable parallel matrix multiplication (as well as solving many other important problems) on distributed memory systems, both theoretically and practically.     

Scalable Parallel Genetic Algorithms

Genetic algorithms, search algorithms based on the genetic processes observed in natural evolution, have been used to solve difficult problems in many different disciplines. When applied to very large-scale problems, genetic algorithms exhibit high computational cost and degradation of the quality of the solutions because of the increased complexity. One of the most relevant research trends in genetic algorithms is the implementation of parallel genetic algorithms with the goal of obtaining quality of solutions efficiently. This paper first reviews the state-of-the-art in parallel genetic algorithms. Parallelization strategies and emerging implementations are reviewed and relevant results are discussed. Second, this paper discusses important issues regarding scalability of parallel genetic algorithms.     

量子机器学习综述

近年来, 机器学习一直是被关注和探讨的研究热点, 被应用到各领域并在其中起着重要作用. 但随着数据量的不断增加, 机器学习算法训练时间越来越长. 与此同时, 量子计算机表现出强大的运算能力. 因此, 有研究人员尝试用量子计算的方法解决机器学习训练时间长的问题, 量子机器学习这一领域应运而生. 量子主成分分析、量子支持向量机、量子深度学习等量子机器学习算法相继被提出, 并有实验证明了量子机器学习算法有显著的加速效果, 使得量子机器学习的研究展现出逐步走高的趋势. 对量子机器学习算法进行综述. 首先介绍量子计算基础; 然后对量子监督学习、量子无监督学习、量子半监督学习、量子强化学习以及量子深度学习5类量子机器学习算法进行介绍; 接着对量子机器学习的相关应用进行介绍并给出了算法实验; 最后进行总结和展望.     

并行LLL算法研究综述

Lenstra-Lenstra-Lovasz（LLL）格基约化算法自1982年被提出以来，已被成功应用于计算机代数、编码理论、密码分析、算法数论、整数规划等众多领域。经过三十多年的发展，串行LLL算法的理论分析和实际效率都已得到显著改进，但仍不能满足密码分析等领域处理较大规模问题的需要。因此，并行LLL算法研究被寄予厚望。对并行LLL算法的研究现状进行了综述，总结了当前并行LLL算法设计与分析中存在的问题和难点，并对其未来发展趋势进行了展望。     

复杂并行机调度问题基于分解的优化算法

针对纺织生产过程中广泛存在的带特殊工艺约束的大规模并行机调度问题，提出了一种基于分解的优化算法。首先将原调度问题分解为机台选择和工件排序两个子问题，然后针对机台选择子问题提出一种进化规划算法，并采用一种具有多项式时间复杂度的最优算法求解工件排序子问题，以得到问题特征信息（即每台机器对应拖期工件数的最小值），该问题特征信息用以指导进化规划算法的迭代过程。不同规模并行机调度问题的数值计算结果及实际制造企业应用效果表明，本文提出的算法是有效的。