基于FILTER算子匹配的增量式DAG计算复用方法

伴随着数据的爆炸式增长,越来越多的大数据业务分析与处理选择分布式计算平台。目前针对大数据的分布式计算框架都支持DAG式的任务编排。由于大数据采集来源以及分布式存储系统特点,很多使用DAG框架进行计算的应用都是增量式的大数据集,但现有的DAG框架对这样的数据集进行计算时有许多冗余,造成计算资源浪费。提出了在DAG框架上进行增量式复用的方法,并针对FILTER算子特点提出了基于FILTER算子匹配的间接复用机制。

基于复用计算的大纹理实时合成

文中提出一种基于复用计算的纹理合成方法,逐步地利用已合成的部分纹理来生成更大的纹理块,以进行后续的纹理合成计算.由此,该方法可节省大量耗时的纹理块选择及缝合计算,提高了合成效率.实验表明,新方法可实时合成2048×2048像素的大纹理,而已有工作至多只能以交互的速度进行这样的合成.

基于复用计算的三维肝脏内部可视化方法研究

虚拟肝脏手术是虚拟手术中的重要内容之一。提出了一种虚拟肝脏建模和体纹理合成方法,通过对肝脏切片数据集的预处理,样条拟合构建虚拟肝脏模型,将基于复用计算的纹理合成扩展到三维空间,应用到构建好的肝脏体上。结果表明,该方法构建的肝脏体能够较好地模拟真实的肝脏结构,绘制的肝脏软组织体纹理具有较好的合成效率和质量,对后续虚拟肝脏手术个性化研究具有一定指导意义。

计算机维护类软件测试用例的设计及复用研究

测试用例的设计是整个测试活动的核心部分,一个好的可复用测试用例可以帮助团队在最少的人力、最快的时间内达到最佳的测试效果。本文针对计算机维护一类软件的特点,研究和设计出可复用的测试用例模型,以在同类软件中提高复用的机会与复用的效果。本文所提出的可复用的测试用例模型是基于测试用例的分类方法、描述方法以及用例库的创建和维护的角度做出的研究。

一个科学计算服务网格应用系统的设计与实现

为扩大计算能力的共享规模,并充分复用已有的计算程序,提出了一个SCS服务的概念.一个SCS服务把一个科学计算程序和它的宿主环境(计算节点)结合起来,构成了一个提供科学计算服务的基本单元.设计了一个网格应用系统(PCG系统)来管理SCS服务.PCG系统是一个3层的架构,顶部是一个Web门户,中间是SCS服务管理器,网格中间件系统位于后端.SCS服务管理器是整个系统的核心部分,包含了四个模块:通讯模块(comm)、需求转换模块(UserRequirementsTranslator)、服务信息查询协商模块(ServiceQuery&Broker)和服务提交管理模块(ServiceHandler).基于GlobusToolkit作为中间件,PCG系统已经在北京大学校园网内跨越多个局域网进行了实现.在PCG系统上,已经成功地运行了若干个实际科学计算的实例.

面向多任务处理的神经网络加速器设计

本文从多任务系统以及算法硬件协同设计的角度出发,提出了一种面向多任务处理的神经网络加速器架构。在算法方面,提出了一套多任务处理算法,能够有效挖掘多任务间的特征重复性和计算重用性;在架构设计方面,设计了能够利用网络参数共享实现多任务间计算重用的加速器架构。该架构采用基于时间片的调度方法,实现了多视觉任务的高效并行处理。实验结果表明,与目前主流的移动图形处理器(GPU)相比,本文提出的加速器能够在多任务推理阶段减少65.80%的乘法运算量,同时获得了 11倍的性能提升。

基于复用计算全息元件的多阶模式偏置波前传感器

提出利用复用计算全息元件实现多阶模式同步的偏置模式波前传感器。对基于复用计算全息元件的多阶模式偏置波前传感器进行了理论分析。为了实现复用计算全息元件,分别编码设计了包含4阶,10阶和20阶Zernike像差模式的复用计算全息图,讨论了复用计算全息图设计中几个关键问题。数值模拟研究了复用计算全息图对单阶和多阶像差模式的探测性能。结果表明,在一定探测范围内,传感器能够响应与全息图内预先记录像差模式相同的待测像差模式,且每一个模式的灵敏度都不同。全息图内记录模式数较少时,单阶模式探测的灵敏度较大;记录模式数越多,模式之间擦除效应越严重,单阶和多阶模式灵敏度都受到一定影响。

基于构件/构架整体叶轮IUMCAM系统的研制

通过对计算机辅助制造(CAM)专项领域可复用构件/构架的研究,针对整体叶轮CAM软件,设计了一个功能划分合理、复用性较强的构件/构架体系模型IUMCAM,从而可以大幅度地降低开发成本,提高软件的生产率以及软件质量。

关于采用DWDM技术传输多路L-Band信号在广播电视领域应用的探讨

高清节目素材的质量要求高、传输时占用的带宽资源大,本文通过对基于密集波分复用技术的多路L-Band信号传输问题的讨论,阐明在信号的光传输领域,采用密集波分技术可以实现有限的资源条件下的高质高效的节目传输。

Cryptanalysis of a Cryptosystem with Non- Commutative Platform Groups

A cryptosystem with non-commutative platform groups based on conjugator search problem was recently introduced at Neural Computing and Applications 2016. Its versatility was illustrated by building a public-key encryption scheme. We propose an algebraic key-recovery attack in the polynomial computational complexity. Furthermore, we peel off the encryption and decryption process and propose attack methods for solving the conjugator search problem over the given non-abelian group. Finally, we provide corresponding practical attack examples to illustrate the attack methods in our cryptanalysis, and provide some improved suggestions.

A Symbol Timing Recovery Algorithm of M-PSK Signals for Burst Modem Applications with Small Packet Size

Feedforward symbol timing recovery techniques are particularly important for initial acquisition in burst modems. However, these techniques either have large calculation burden or sensitive to frequency offsets. In this paper, we proposed an efficient symbol timing recovery algorithm of MPSK signals named OMQ(Ordered Maximum power using Quadratic approximation partially) algorithm which is based on the Quadratic Approximation(QA) algorithm. We used ordered statistic sorting method to reduce the computational complexity further, meanwhile maximum mean power principle was used to decrease frequency offset sensitivity. The proposed algorithm adopts estimation-down sampling structure which is suitable for small packet size transmission. The results show that, while comparing with the QA algorithm, the computational complexity is reduced by 75% at most when 8 samples per symbol are used. The proposed algorithm shows better performance in terms of the jitter variance and sensitivity to frequency offsets.

Interference Alignment Based on Subspace Tracking in MIMO Cognitive Networks with Multiple Primary Users

The interference alignment （IA） algorithm based on FDPM subspace tracking （FDPM-ST IA） is proposed for MIMO cognitive network （CRN） with multiple primary users in this paper. The feasibility conditions of FDPM-ST IA is also got. Futherly, IA scheme of secondary network and IA scheme of primary network are given respectively without assuming a priori knowledge of interference covariance matrices. Moreover, the paper analyses the computational complexity of FDPM-ST IA. Simulation results and theoretical calculations show that the proposed algorithm can achieve higher sum rate with lower computational complexity.

A Novel Two-step Frequency Offset Estimator for OFDM

Based on an orthogonal frequency division multiplexing(OFDM) training symbol with L identical parts, a novel carrier frequency offset (CFO) estimator is proposed for OFDM systems. The CFO is estimated in two steps, fine estimate and coarse estimate. In the first step, the fine estimation is performed based on the principle of minimum variance. However, the fine estimation has ambiguity since its estimate range is limited. In the second step, the coarse estimation is obtained, which results in a larger estimate range but less precision. Using the coarse estimation, the ambiguity of fine estimation is resolved. To fully use the correlation among L identical parts, the fine estimation resolved the ambiguity and the coarse estimation are optimally combined to obtain the final estimation. Furthermore, the estimation variance of the proposed method is derived. Simulation results demonstrate that the novel two-step estimator outperforms the conventional two-step estimator in terms of estimate performance and computational complexity.

Numerical study and acceleration of LBM-RANS simulation of turbulent flow

The coupled models of LBM （Lattice Boltzmann Method） and RANS （Reynolds-Averaged Navier-Stokes） are more practical for the transient simulation of mixing processes at large spatial and temporal scales such as crude oil mixing in large-diameter storage tanks. To keep the efficiency of parallel computation of LBM, the RANS model should also be explicitly solved; whereas to keep the numerical stability the implicit method should be better for PANS model. This article explores the numerical stability of explicit methods in 2D cases on one hand, and on the other hand how to accelerate the computation of the coupled model of LBM and an implicitly solved RANS model in 3D cases. To ensure the numerical stability and meanwhile avoid the use of empirical artificial lim- itations on turbulent quantities in 2D cases, we investigated the impacts of collision models in LBM （LBGK, MRT） and the numerical schemes for convection terms （WENO, TVD） and production terms （FDM, NEQM） in an explic- itly solved standard k-e model. The combination of MRT and TVD or MRT and NEQM can be screened out for the 2D simulation of backward-facing step flow even at Re = 107. This scheme combination, however, may still not guarantee the numerical stability in 3D cases and hence much finer grids are required, which is not suitable for the simulation of industrial-scale processes.Then we proposed a new method to accelerate the coupled model of LBM with RANS （implicitly solved）. When implemented on multiple GPUs, this new method can achieve 13.5-fold accelera- tion relative to the original coupled model and 40-fold acceleration compared to the traditional CFD simulation based on Finite Volume （FV） method accelerated by multiple CPUs. This study provides the basis for the transient flow simulation of larger spatial and temporal scales in industrial applications with LBM-RANS methods.

Joint Timing and Frequency Synchronization for OFDM based on Fractional Fourier Transform

In this paper a joint timing and frequency synchronization method based on Fractional Fourier Transform （FIFT） is proposed for Orthogonal Frequency-Division Multiplexing （OFDM） system. The combination of two chirp signals with opposite chirp rates are used as the training signal, the received training signal with timing and frequency offset is transformed by FrFT and the two peaks representing two chirps in FrFT domain are detected, then the position coordinates of the two peaks are precisely corrected and substituted into an equation group to calculate timing and frequency offset simultaneously. This method only needs one FrFT calculation to implement synchronization, the computational complexity is equal to that of FFT and less than that of correlation or maximum likelihood calculation of existing methods, and estimation range of frequency offset is Large, greater than half the signal bandwidth, while the simulation results show that even at low SNR it can accurately estimate timing and frequency offset and the estimation error is less than that of existing methods.

ADAPTIVE BLIND ESTIMATION ALGORITHM FOR OFDM-MIMO RADIO SYSTEMS OVER MULTIPATH CHANNELS

This paper investigates the blind algorithm for channel estimation of Orthogonal Frequency Division Multiplexing-Multiple Input Multiple Output (OFDM-MIMO) wireless communication system using the subspace decomposition of the channel received complex baseband signals and proposes a new two-stage blind algorithm. Exploited the second-order cyclostationarity inherent in OFDM with cyclic prefix and the characteristics of the phased antenna, the practical HIPERLAN/2 standard based OFDM-MIMO simulator is established with the sufficient consideration of statistical correlations between the multiple antenna channels under wireless wideband multipath fading environment, and a new two-stage blind algorithm is formulated using rank reduced subspace channel matrix approximation and adaptive Constant Modulus (CM)criterion. Simulation results confirm the theoretical analysis and illustrate that the proposed algorithm is capable of tracking matrix channel variations with fast convergence rate and improving acceptable overall system performance over various common wireless and mobile communication links.

Performance of a novel carrier frequency offset estimation algorithm for OFDM-based WLANs

This paper presents a novel carrier frequency offset estimation (CFO) algorithm for orthogonal frequency division multiplexing (OFDM)-based Wireless Local Area Networks (WLANs). Compared with previous approaches, this paper extends the whole frequency offset acquisition range by embedding a synthetic algorithm according to the preamble structure of WLANs symbols. The numerical results presented support the effectiveness of this algorithm by which the estimation error of the whole carrier frequency offset in the WLANs is effectively decreased.

An efficient channel estimator for OFDM system with sparse multipath fading

A channel estimator used in sparse muhipath fading channel for orthogonal frequency division multiplexing （OFDM） system is proposed. The dimension of signal subspace can be reduced to improve the performance of channel estimation. The simplified version of original subspace fitting algorithm is employed to derive the sparse multipaths. In order to overcome the difficulty of termination condition, we consider it as a model identification problem and the set of nonzero paths is found under the generalized Akaike information criterion （GAIC）. The computational complexity can be kept very low under proper training design. Our proposed method is superior to other related schemes due to combining the procedure of selecting the most probable taps with GAIC model selection. Simulation in hilly terrain （HT） channel shows that the proposed method has an outstanding performance.

A novel improved SLM scheme of the PAPR reduction technology in CO-OFDM systems

The higher peak-to-average power ratio(PAPR) is a major shortcoming of coherent optical orthogonal frequency division multiplexing(CO-OFDM) systems. Selective mapping(SLM) technology can effectively reduce the probability of high PAPR, but it has higher computational complexity, and requires additional bandwidth to transmit the side information, which will affect the transmission efficiency of the system. In response to these shortcomings, a novel improved SLM(NI-SLM) scheme with low complexity and without side information is proposed. Simulation results show that the proposed scheme can exponentially reduce the computational complexity, and the bit error rate(BER) performance can greatly approach the original signal. What's more, it shows the better PAPR reduction performance.

Computing Sparse GCD of Multivariate Polynomials via Polynomial Interpolation

The problem of computing the greatest common divisor（GCD） of multivariate polynomials, as one of the most important tasks of computer algebra and symbolic computation in more general scope, has been studied extensively since the beginning of the interdisciplinary of mathematics with computer science. For many real applications such as digital image restoration and enhancement,robust control theory of nonlinear systems, L1-norm convex optimization in compressed sensing techniques, as well as algebraic decoding of Reed-Solomon and BCH codes, the concept of sparse GCD plays a core role where only the greatest common divisors with much fewer terms than the original polynomials are of interest due to the nature of problems or data structures. This paper presents two methods via multivariate polynomial interpolation which are based on the variation of Zippel＇s method and Ben-Or/Tiwari algorithm, respectively. To reduce computational complexity, probabilistic techniques and randomization are employed to deal with univariate GCD computation and univariate polynomial interpolation. The authors demonstrate the practical performance of our algorithms on a significant body of examples. The implemented experiment illustrates that our algorithms are efficient for a quite wide range of input.