基于GPU计算的并行PLS算法研究与实现

偏最小二乘算法(PLS)是与红外、近红外光谱分析结合使用最为广泛的化学计量学算法,然而当前PLS算法通常采用单线程方式实现,当校正模型数量多或样本数量大、波长点数和主成分数较多,模型需对光谱预处理和波长选择方法反复优化时,计算十分缓慢。为大幅提高建模速度,该文提出了一种基于图形处理器(GPU)的并行计算策略,利用具有大规模并行计算特性的GPU作为计算设备,结合CUBLAS库函数实现了基于GPU并行的PLS建模算法(CUPLS)。利用近红外光谱数据集进行性能对比实验,结果表明CUPLS建模算法较传统单线程实现的PLS算法,加速比可达近42倍,极大地提升了化学计量学算法的建模效率。该方法亦可用于其它化学计量学算法的加速。     

Distributed localized bi-objective search

We propose a new distributed heuristic for approximating the Pareto set of bi-objective optimization problems. Our approach is at the crossroads of parallel cooperative computation, objective space decomposition, and adaptive search. Given a number of computing nodes, we self-coordinate them locally, in order to cooperatively search different regions of the Pareto front. This offers a trade-off between a fully independent approach, where each node would operate independently of the others, and a fully centralized approach, where a global knowledge of the entire population is required at every step. More specifically, the population of solutions is structured and mapped into computing nodes. As local information, every node uses only the positions of its neighbors in the objective space and evolves its local solution based on what we term a ‘localized fitness function  ’. This has the effect of making the distributed search evolve, over all nodes, to a high quality approximation set, with minimum communications. We deploy our distributed algorithm using a computer cluster of hundreds of cores and study its properties and performance on ρ$\rho$MNK-landscapes. Through extensive large-scale experiments, our approach is shown to be very effective in terms of approximation quality, computational time and scalability.     

Resolvability of infinite designs

In this paper we examine the resolvability of infinite designs. We show that in stark contrast to the finite case, resolvability for infinite designs is fairly commonplace. We prove that every t  -(v,k,Λ)$(v,k,\Lambda )$ design with t finite, v   infinite and k,λ<v$k,\lambda <v$ is resolvable and, in fact, has α   orthogonal resolutions for each α<v$\alpha <v$. We also show that, while a t  -(v,k,Λ)$(v,k,\Lambda )$ design with t and λ finite, v   infinite and k=v$k=v$ may or may not have a resolution, any resolution of such a design must have v parallel classes containing v   blocks and at most λ−1$\lambda -1$ parallel classes containing fewer than v   blocks. Further, a resolution into parallel classes of any specified sizes obeying these conditions is realisable in some design. When k<v$k<v$ and λ=v$\lambda =v$ and when k=v$k=v$ and λ is infinite, we give various examples of resolvable and non-resolvable t  -(v,k,Λ)$(v,k,\Lambda )$ designs.     

Some results on series and parallel systems of randomized components

We study the system (series/parallel) where the components are randomly chosen from two different batches. We assume that one batch is more reliable than the other in some stochastic sense. In the case of series systems we show that, under certain conditions, lifetime of one system dominates that of the other in different stochastic orders viz. hazard rate, down shifted hazard rate and likelihood ratio orders. Further, we show that the same principle holds for the reversed hazard rate and the likelihood ratio orders in the case of parallel systems.     

一种基于大数据处理的异构私有云系统

云计算和大数据已成为IT领域的研究热点,如何将云计算在数据存储和数据处理方面的优势应用于大数据领域具有重要的实际应用价值.开源的云平台OpenStack可方便地从硬件管理方面构建私有云,其存储模块Swift能够支持PB级的大数据存储.开源的云平台Hadoop在数据处理方面具有很强的优势,但在支持超大数据存储方面存在不足.通过对OpenStack中的存储模块Swift和Hadoop中的文件处理模块HDFS的比较分析,提出了将Swift和Hadoop的MapReduce技术结合来构建企业处理大数据的私有云计算系统方案.分析结果显示该方案是可行的,这种异构的私有云系统可以整合不同云计算平台各自的优势进行高效的大数据处理.     

基于云计算的信息服务通用模型研究

概括了云计算技术的普遍特征,分析了云计算技术在信息服务业的典型应用案例及其局限性,继而针对性得提出了一种基于云计算的信息服务通用模型,并通过量化分析论证了该模型的可行性.     

Computation rate maximization for IRS-enhanced dynamic spectrum access assisted MEC system

Intelligent reflecting surface (IRS)-enhanced dynamic spectrum access (DSA) is a promising technology to enhance the performance of the mobile edge computing (MEC) system. In this paper, we consider the integration of the IRS enhanced DSA technology to a MEC system, and study the pertinent joint optimization of the phase shift coefficients of the IRS, the transmission powers, the central processing unit (CPU) frequencies, as well as the task offloading time allocations of the secondary users (SUs) to maximize the average computation bits of the SUs. Due to the non-convexity, the formulated problem is difficult to solve. In order to tackle this difficulty, we decompose the optimization problem into tractable subproblems and propose an alternating optimization algorithm to optimize the optimization variables in an iterative fashion. Numerical results are provided to show the effectiveness and the correctness of the proposed algorithm.     

Applying Parallel Factor Analysis models to HPLC diode array detector datasets reveals strain dependent regulation of polyketide biosynthesis in Fusarium graminearum, Fusarium culmorum and Fusarium pseudograminearum

Fungal secondary metabolites confer an indiscriminate resource of bioactive compounds with both pharmaceutical and industrial relevance. Synthesis of these compounds is tightly controlled through regulatory networks. These networks act in response to the environmental conditions in which the fungus grows. One class of fungal secondary metabolites that have drawn increased attention, is the polyketides produced by a high number of fungi and bacteria. The increasing number of publicly available genomic sequences of filamentous fungi, combined with new and highly efficient DNA-cloning techniques, has allowed the use of new efficient approaches to study the regulation and synthesis of these compounds. With the availability of new molecular genetic tools, a new bottleneck has occurred in fungal molecular biology; the analytical chemical analysis of the generated transgenic mutants. In this study, a method based on PARAFAC data analysis of HPLC data is presented and shown useful for gaining overview of complex and diverse datasets. Further, the strain specific and nutrient dependent regulation of polyketide synthesis is discussed.     

Three-component one-pot synthetic route to 2-amino-5-alkylidene-thiazol-4-ones

A fast and straightforward three-component reaction to 2-amino-5-alkylidene-thiazol-4-ones is described. The one-pot methodology, reported for the first time, involves Knoevenagel condensation of aromatic aldehydes and rhodanine followed by displacement of the thiocarbonyl sulfur with primary or secondary amines in the same reaction mixture. The reactions were performed using a dedicated microwave reactor, which enabled short reaction times and easy work-up.     

Parallel Fock matrix construction program for molecular orbital calculation--specific computer with a hierarchical network

A parallel Fock matrix construction program for a hierarchical network has been developed on the molecular orbital calculation-specific EHPC system. To obtain high parallelization efficiency on the hierarchical network system, a multilevel dynamic load-balancing scheme was adopted, which provides equal load balance and localization of communications on a tree-structured hierarchical network. The parallelized Fock matrix construction routine was implemented into a GAMESS program on the EHPC system, which has a tree-structured hierarchical network. Benchmark results on a 63-processor system showed high parallelization efficiency even on the tree-structured hierarchical network.