nPME对GROMACS软件并行计算性能的影响分析

在分子动力学并行计算的过程中,正确地处理好并行规模与PME(Particle-Mesh Ewald)方法的任务分配,对于提高分子动力学的并行效率具有非常重要的影响。以常用的分子动力学软件Gromacs[1-3]为例,利用上海超级计算中心"魔方"超级计算平台,就不同并行规模与不同PME任务分配方式下的计算性能进行测试。发现并行能在一定的范围内显著的提高Gromacs的计算效率,但是当并行规模超过一定值后,计算效率反而下降。另一方面,当并行规模较小时,不设置PME专属节点的计算效率较高;但当并行规模较大时,合理的设置PME专属节点,能显著地提高计算的效率。这一结果能对从事分子动力学模拟的研究人员合理分配并行资源提供帮助。     

pest: Fast approximate keyword search in semantic data using eigenvector-based term propagation

We present pest, a novel approach to the approximate querying of graph-structured data such as RDF that exploits the data's structure to propagate term weights between related data items. We focus on data where meaningful answers are given through the application semantics, e.g., pages in wikis, persons in social networks, or papers in a research network such as Mendeley. The pest matrix generalizes the Google Matrix used in PageRank with a term-weight dependent leap and accommodates different levels of (semantic) closeness for different relations in the data, e.g., friend vs. co-worker in a social network. Its eigenvectors represent the distribution of a term after propagation. The eigenvectors for all terms together form a (vector space) index that takes the structure of the data into account and can be used with standard document retrieval techniques. In extensive experiments including a user study on a real life wiki, we show how pest improves the quality of the ranking over a range of existing ranking approaches, yet achieves a query performance comparable to a plain vector space index.     

Community detection in complex networks: Multi-objective discrete backtracking search optimization algorithm with decomposition

Community detection is believed to be a very important tool for understanding both the structure and function of complex networks, and has been intensively investigated in recent years. Community detection can be considered as a multi-objective optimization problem and the nature-inspired optimization techniques have shown promising results in dealing with this problem. In this study, we present a novel multi-objective discrete backtracking search optimization algorithm with decomposition for community detection in complex networks. First, we present a discrete variant of the backtracking search optimization algorithm (DBSA) where the updating rules of individuals are redesigned based on the network topology. Then, a novel multi-objective discrete method (MODBSA/D) based on the proposed discrete variant DBSA is first proposed to minimize two objective functions in terms of Negative Ratio Association (NRA) and Ratio Cut (RC) of community detection problems. Finally, the proposed algorithm is tested on some real-world networks to evaluate its performance. The results clearly show that MODBSA/D has effective and promising performance for dealing with community detection in complex networks.     

Categorical data fuzzy clustering: An analysis of local search heuristics

The fuzzy c partition of a set of qualitative data is the problem of selecting the optimal c   centroids that are the most representative of the whole population. Moreover, a set of weights _wij$w_{\mathit{ij}}$ must be determined, describing the fuzzy membership function of pattern i to the cluster represented by centroid j. Both problems are formulated by a single mathematical programming problem, that is an extension of the classic p-median models often used for clustering. The new objective function is neither concave nor convex and the application requires the clustering of many thousands of data, therefore heuristic methods are to be developed to find the best fuzzy partition.     

Hybrid genetic algorithm with adaptive local search scheme

This paper proposes a hybrid genetic algorithm (a-hGA) with adaptive local search scheme. For designing the a-hGA, a local search technique is incorporated in the loop of genetic algorithm (GA), and whether or not the local search technique is used in the GA is automatically determined by the adaptive local search scheme. Two modes of adaptive local search schemes are developed in this paper. First mode is to use the conditional local search method that can measure the average fitness values obtained from the continuous two generations of the a-hGA, while second one is to apply the similarity coefficient method that can measure a similarity among the individuals of the population of the a-hGA. These two adaptive local search schemes are included in the a-hGA loop, respectively. Therefore, the a-hGA can be divided into two types: a-hGA1 and a-hGA2. To prove the efficiency of the a-hGA1 and a-hGA2, a canonical GA (cGA) and a hybrid GA (hGA) with local search technique and without any adaptive local search scheme are also presented. In numerical example, all the algorithms (cGA, hGA, a-hGA1 and a-hGA2) are tested and analyzed. Finally, the efficiency of the proposed a-hGA1 and a-hGA2 is proved by various measures of performance.     

Efficient data structures for backtrack search SAT solvers

The implementation of efficient Propositional Satisfiability (SAT) solvers entails the utilization of highly efficient data structures, as illustrated by most of the recent state-of-the-art SAT solvers. However, it is in general hard to compare existing data structures, since different solvers are often characterized by fairly different algorithmic organizations and techniques, and by different search strategies and heuristics. This paper aims the evaluation of data structures for backtrack search SAT solvers, under a common unbiased SAT framework. In addition, advantages and drawbacks of each existing data structure are identified. Finally, new data structures are proposed, that are competitive with the most efficient data structures currently available, and that may be preferable for the next generation SAT solvers.     

Efficient data structures for backtrack search SAT solvers

The implementation of efficient Propositional Satisfiability (SAT) solvers entails the utilization of highly efficient data structures, as illustrated by most of the recent state-of-the-art SAT solvers. However, it is in general hard to compare existing data structures, since different solvers are often characterized by fairly different algorithmic organizations and techniques, and by different search strategies and heuristics. This paper aims the evaluation of data structures for backtrack search SAT solvers, under a common unbiased SAT framework. In addition, advantages and drawbacks of each existing data structure are identified. Finally, new data structures are proposed, that are competitive with the most efficient data structures currently available, and that may be preferable for the next generation SAT solvers.     

用哈希算法和二分法实现中文词典的快速查找

分词词典是汉语自动分词系统中的一个基本组成部分,其查询速度直接影响到分词系统的处理速度。文章提出并实现了一种用哈希算法和二分查找算法相结合的中文单词查找算法,实验显示,该算法可以实现对字符串的快速查找。     

A flabellate overlay network for multi-attribute search

Peer-to-peer (P2P) technology provides a popular way of distributing resources, sharing, and locating in a large-scale distributed environment. However, most of the current existing P2P systems only support queries over a single resource attribute, such as file name. The current multiple resource attribute search methods often encounter high maintenance cost and lack of resilience to the highly dynamic environment of P2P networks. In this paper, we propose a Flabellate overlAy Network (FAN), a scalable and structured underlying P2P overlay supporting resource queries over multi-dimensional attributes. In FAN, the resources are mapped into a multi-dimensional Cartesian space based on the consistent hash values of the resource attributes. The mapping space is divided into non-overlapping and continuous subspaces based on the peer’s distance. This paper presents strategies for managing the extended adjacent subspaces, which is crucial to network maintenance and resource search in FAN. The algorithms of a basic resource search and range query over FAN are also presented in this paper. To alleviate the load of the hot nodes, a virtual replica network (VRN) consisting of the nodes with the same replicates is proposed for replicating popular resources adaptively. The queries can be forwarded from the heavily loaded nodes to the lightly loaded ones through VRN. Theoretical analysis and experimental results show that FAN has a higher routing efficiency and lower network maintenance cost over the existing multi-attribute search methods. Also, VRN efficiently balances the network load and reduces the querying delay in FAN while invoking a relatively low overhead.     

网络爬虫在Web信息搜索与数据挖掘中应用

分析了万维网不良网络信息对网络文化安全带来的挑战,提出了Web信息搜索与数据挖掘体系结构,并介绍了该体系结构中的关键技术和运行原理.分析了普通爬虫所实现的功能和不足之后,重点论述了该爬虫的工作原理、实现方式和性能分析以及该爬虫不同于其它爬虫的功能和在Web信息搜索与数据挖掘体系中应用.通过试验测试表明,该爬虫能够很好地获取万维网上的各种信息资源,有助于网络文化内容监测与管理.     

Aggregated search of data and services

From a user perspective, data and services provide a complementary view of an information source: data provide detailed information about specific needs, while services execute processes involving data and returning an informative result as well. For this reason, users need to perform aggregated searches to identify not only relevant data, but also services able to operate on them. At the current state of the art such aggregated search can be only manually performed by expert users, who first identify relevant data, and then identify existing relevant services.     

Fast exact k nearest neighbors search using an orthogonal search tree

The problem of k nearest neighbors (kNN) is to find the nearest k neighbors for a query point from a given data set. In this paper, a novel fast kNN search method using an orthogonal search tree is proposed. The proposed method creates an orthogonal search tree for a data set using an orthonormal basis evaluated from the data set. To find the kNN for a query point from the data set, projection values of the query point onto orthogonal vectors in the orthonormal basis and a node elimination inequality are applied for pruning unlikely nodes. For a node, which cannot be deleted, a point elimination inequality is further used to reject impossible data points. Experimental results show that the proposed method has good performance on finding kNN for query points and always requires less computation time than available kNN search algorithms, especially for a data set with a big number of data points or a large standard deviation.     

Visual and memory search in complex environments: determinants of eye movements and search performance

Previous research on visual and memory search revealed various top down and bottom up factors influencing performance. However, utilising abstract stimuli (e.g. geometrical shapes or letters) and focussing on individual factors has often limited the applicability of research findings. Two experiments were designed to analyse which attributes of a product facilitate search in an applied environment. Participants scanned displays containing juice packages while their eye movements were recorded. The familiarity, saliency, and position of search targets were systematically varied. Experiment 1 involved a visual search task, whereas Experiment 2 focussed on memory search. The results showed that bottom up (target saliency) and top down (target familiarity) factors strongly interacted. Overt visual attention was influenced by cultural habits, purposes, and current task demands. The results provide a solid database for assessing the impact and interplay of fundamental top down and bottom up determinants of search processes in applied fields of psychology. Practitioner Summary: Our study demonstrates how a product (or a visual item in general) needs to be designed and placed to ensure that it can be found effectively and efficiently within complex environments. Corresponding product design should result in faster and more accurate visual and memory based search processes.