基于任务映射的并行辐射度算法

基于辐射度算法的大规模数据场景的可视化模拟中所面临的主要问题是其计算时间无法满足实时性的要求,提出了任务映射模型较好地解决了各计算节点的负载均衡性要求,从而有效提高辐射度算法的并行计算效率。基于该模型开发了并行辐射度算法并在超级计算机模拟器上进行了测试,试验结果表明,该算法具有良好的可扩展性和较高的计算加速比。     

基于递进卷积神经网络的台标识别及其并行化

针对台标的视觉特征,提出一种基于递进卷积神经网络的台标识别算法.该网络不仅有对图像特征进行隐性提取的卷积层和采样层,还包括识别常规台标的泛化模块和识别偏差台标的特异模块.针对串行卷积神经网络训练耗时长的缺点,提出基于Spark的并行递进卷积神经网络算法,采用数据共享及批处理方式对算法模型进行并行化处理.实验证明,递进卷积神经网络算法对台标进行识别能达到98％的正确率,多节点并行化卷积神经网络相比于单节点模型能有效缩短80％以上训练所需的时间.     

并行K均值聚类和贪婪算法融合的软硬件划分

提出了一种并行K均值聚类和贪婪算法融合的软硬件划分算法.算法首先将有相似属性的任务节点通过并行K均值聚类算法组成一个大的任务节点,而后使用贪婪算法划分由大的任务节点组成的系统.实验结果表明,这种软硬件划分的方法具有高效率及高面积利用率的特点,尤其对多节点的复杂任务算法的求解时间远小于贪婪算法.     

Multiuser subcarriers and bit allocation combining adaptive mapping for SC-FDMA system

In this article, a multiuser single carrier frequency division multiple access （SC-FDMA） system is considered, based on which an adaptive subcarrier and bit allocation algorithm is investigated. The algorithm has been used to achieve a subcarrier mapping mode in this system, which combines the advantages of single- and multi-carrier transmissions, such as, low peak to average power ratio, orthogonality of signals of different users, and low complexity. Simulation results show that it has a similar performance as that of the adaptive allocation algorithm in the orthogonal frequency division multiple （OFDM） system and the proposed mapping mode has a performance gain over the two existing mapping modes at the link level.     

基于量子布谷鸟搜索的认知无线网络频谱分配

为了有效解决认知无线网络频谱分配的离散优化问题,将量子计算引入布谷鸟搜索算法,提出了一种新的组合优化算法——量子布谷鸟搜索算法。该算法使用量子鸟窝表征问题的多维解,通过Lévy flights随机游动方式和量子突变策略快速搜索到全局最优位置。通过使用基准函数验证了算法的高效性,并提出了一种基于量子布谷鸟搜索的认知无线网络频谱分配方法。然后与经典频谱分配算法在不同的网络效益函数下进行仿真性能比较。结果表明,所提出的频谱分配方法能够较快找到全局最优解,并且在不同网络效益函数下均优于已有的经典频谱分配算法。     

A tree‐based algorithm for virtual infrastructure allocation with joint virtual machine and network requirements

Cloud providers have introduced the on‐demand provisioning of virtual infrastructures (VIs) to deliver virtual networks of computing resources as a service. By combining network and computing virtualization, providers allow traffic isolation between hosted VIs. Taking advantage of this opportunity, tenants have deployed private VIs with application‐optimized network topologies to increase quality of experience of final users. One of the main open challenges in this scenario is the allocation of physical resources to host VIs in accordance with quality of service computing (eg, virtual CPUs and memory) and network requirements (guaranteed bandwidth and specific network topology). Moreover, a VI can be allocated anywhere atop a network datacenter, and because of its NP‐hard complexity, the search for optimal solutions has a limited applicability in cloud providers as requesting users seek an immediate response. The present work proposes an algorithm to accomplish the VI allocation by applying tree‐based heuristics to reduce the search space, performing a joint allocation of computing and network resources. So as to accomplish this goal, the mechanism includes a strategy to convert physical and virtual graphs to trees, which later are pruned by a grouped accounting algorithm. These innovations reduce the number of comparisons required to allocate a VI. Experimental results indicate that the proposed algorithm finds an allocation on feasible time for different cloud scenarios and VI topologies, while maintaining a high acceptance rate and a moderate physical infrastructure fragmentation.     

机群系统上基于映射和抽样划分的并行聚类算法

采用映射和抽样划分方法,基于MPI消息传递编程模式,在机群系统上设计与实现一种并行聚类算法.该算法将生物基因序列映射成整数值,采用整数值取代字符串进行聚类,使得聚类过程快速,通过多次抽样一次聚类寻找初值在一定程度上避免了聚类结果陷入局部解的问题,优化了聚类质量.在PC机群系统上对基因序列进行并行聚类的实验结果表明该算法获得了较好的加速和可扩展性.     

基于图分割的流应用多处理器映射算法

为了充分利用多处理器平台所提供的计算资源,需要将应用以适当的方式映射到不同处理器,从而最大程度地挖掘应用所提供的并发性以满足应用严格的实时性要求。提出了并发图来量化、建模应用任务间的并发性,提出了一种基于自同步调度的并发图构建算法,并将任务映射问题转换成图分割问题,然后将并发图分割问题建模为纯0-1整数线性规划模型并采用ILP求解器获得最优解。采用了大量随机生成的同步数据流图以及一组实际应用对所提方法进行性能评估,实验结果表明所提方法性能优于已有算法。     

Resource allocation based on genetic algorithm for multi-hop OFDM system with non-regenerative relaying

This article investigates resource allocation in multi-hop orthogonal frequency division multiplexing (OFDM) system with amplifying-and-forwarding relaying to maximize the end-to-end capacity. Most existing methods for multi-hop system focus on power allocation or subcarrier selection separately, but joint resource allocation is rarely considered due to the absence of effective interaction schemes. In this work, a novel joint resource allocation methodology is proposed based on Partheno genetic algorithm (PGA), which produces excellent subcarrier allocation set (referred to as individual in PGA) with higher capacity by evolution operator generation by generation. In addition, an adaptive power allocation is also designed to evaluate the fitness of PGA and further enhance the system capacity. Both theoretical analysis and simulated results show the effectiveness of the proposed joint strategy. It outperforms the traditional method by as much as 40% capacity improvement for 3-hop relaying system when system power is high, and obtains much more capacity enhancement percent under conditions of low system power.     

面向分布式环境的信号驱动任务调度算法

为优化IaaS服务的执行效率,提出面向IaaS的信号驱动任务调度算法,该算法根据IaaS模型的结构特征建立控制子系统和节点子系统,根据任务的结构特征建立任务的DAG(directed acyclic graph)调度模型,并建立各任务分片的状态转化机制及控制子系统和节点子系统间的信号通信机制。以系统间信号交互的方式驱动任务分片的状态改变,并在每一调度时刻来临时利用并行优化选择策略分配任务分片。由于本算法采用了模拟IaaS模型的双系统控制方式,使本算法与IaaS模型的分布式体系相兼容且复杂度较低。最后通过实验验证了所提算法的有效性和实用性。     

Resource allocation algorithm for situation awareness based on multiple-choice quadratic knapsack

In order to deal with the potential cyber-threat and improve the security situation by using limited resource properly, the optimal allocation of resource focused on cyber security situation. The coherence of network situation lead to the fact that the enhancement of certain item may also affect some other items, and different amount of investment may also result in different degree of impact, therefore, the problem was extracted into the multiple-choice quadratic knapsack problem. The characteristics of quadratic knapsack problem was used to model the interactions among the situation indi-cator items, meanwhile used the multiple choice knapsack problem to model the multiple investment choice for each item. A branch and bound algorithm was conducted by using the semi-definite relaxation. The experiment results show the ac-curacy and efficiency of proposed algorithm.     

一种基于特征向量的高效虚拟现实辐射度算法

辐射度技术对于具有真实感的图像的合成十分重要。提出了一种基于特征向量的虚拟现实环境中辐射度计算的新方法。该方法采用了并行计算方法计算形状因子矩阵的特征值和特征向量,进而实现对虚拟现实环境中的辐射度进行高效的计算。其性能总体评价结果显示该方法显著降低了复杂环境中集群计算机系统的有效计算时间,提高了计算效率和加速比。其成果可广泛应用在图像处理、人机交互、可视化、虚拟现实等领域。     

基于萤火虫算法的多中继功率分配

为了充分实现中继协作,降低多中继协作通信系统功率分配优化问题的计算复杂度,提出了基于萤火虫算法的多中继功率分配方案。在一定的总功率和节点功率约束下,以最大化平均信噪比为优化目标函数,建立了多中继协作系统的功率分配最优化模型。选取该目标函数作为萤火虫的适应度函数,用向量表示萤火虫的状态,该向量的维数为待分配源节点和中继节点的个数,通过萤火虫聚集得到种群中最好的萤火虫,即可获得渐进最优功率分配。仿真结果表明,与平均功率分配相比,基于萤火虫算法的功率分配方案能降低2.44%~6.17%的比特差错率,提高了系统性能。     

基于选择性子载波的OFDM中继信道无线资源分配算法

针对OFDM中继信道,本文提出了一种基于选择性子载波的无线资源分配算法.文中通过选择信道条件最好的部分子载波进行平均功率分配,并对中继子载波进行优化分配.仿真结果表明,该算法不仅能够提高系统可获取的传输速率,而且可以降低运算复杂度.     

基于行为规律的搜索资源分配新算法

针对具有行为规律的目标搜索问题,提出一种搜索资源分配算法。该方法以目标在各搜索区域的概率分布为基础,利用最优搜索理论分配区域搜索时长;以目标在各搜索区域的行为规律为基础,利用包络检测等方法决定区域开始搜索的时刻。针对网站关键词搜索的实验显示,根据目标行为规律在时间上相关程度的不同,本算法相对于最优搜索算法的性能提升在15%~50%之间,在对大量信息源进行信息搜索时具有应用价值。     

基于多用户MIMO/OFDM系统的空间子信道分配算法

对多用户MIMO/OFDM系统空间子信道进行了研究,以获取最大的系统吞吐量为目标,推导了子载波分配准则,提出了一种自适应空间子信道分配算法,并给出了相应的算法流程。仿真结果表明,该算法具有良好的性能,有效地提高了系统的传输速率。     

基于Logistic映射和z-映射的图像分组加密算法

针对图像加密,提出了一种基于Logistic映射和z-映射的图像分组加密算法,该算法在每一轮迭代中,交替使用矩阵置乱、非线性数值替换和数值关联混迭三级操作。经大量实验验证,该算法能够有效抵抗明文攻击、差分攻击、统计分析,密钥空间较大,因而安全性较高,并且加解密速度相当快,适用于数据存储。     

OFDM系统中的自适应比特功率分配算法

正交频分复用(OFDM)系统各子信道具有不同的衰落特性,如果采用相同的调制方式,则不能更好地利用系统资源。在系统容量受限和误比特率一定的条件下,文章提出了一种高效的OFDM自适应比特功率分配算法。在要求误比特率BER<10-3时,该算法与Chow算法相比,发射总功率减小了2.5 dB;与Hughes-Hartogs算法相比,系统消耗的总时间也明显缩短。仿真结果验证了算法的有效性。     

存在反馈延迟时V-BLAST系统的功率分配算法

针对信道反馈有延迟条件下的垂直贝尔实验室分层空时(Vertical Bell Laboratories Layered Space Time,V-BLAST)系统,提出基于最小均方误差检测(Minimize Mean Square Error,MMSE)的自适应功率分配算法.通过系统建模和性能分析,推导出系统瞬时信干噪比的条件概率密度函数,并得到V-BLAST系统的平均BER表达式.发 送端在总功率约束条件下,以瞬时BER为优化目标,利用拉格朗日极值法求解自适应功率分配矩阵.仿真结果表明,在信道反馈有延迟时,与等功率分配相比,采用所提算法可显著改善系统的BER性能.     

Improved satellite resource allocation algorithm based on DRL and MOP

In view of the multi-objective optimization (MOP) problem of sequential decision-making for resource allocations in multi-beam satellite systems,a deep reinforcement learning(DRL) based DRL-MOP algorithm was proposed to improve the system performance and user satisfaction degree.With considering the normalized weighted sum of spectrum efficiency,energy efficiency,and satisfaction index as the optimization goal,the dynamically changing system environments and user arrival model were built by the proposed algorithm,and the optimization of the accumulative performance in satellite systems based on DRL and MOP was realized.Simulation results show that the proposed algorithm can solve the MOP problem with rapid convergence ability and low complexity,and it is obviously superior to other algorithms in terms of system performance and user satisfaction optimization.