模型参数迁移的改进LS-SVM算法及其应用

针对目标领域样本不足的问题,提出了一种基于模型参数迁移的改进LS-SVM算法.在LS-SVM的基础上加入了迁移项,使得模型既保持了LS-SVM的基本属性,即最大化几何间隔,又能够有效地使用源领域的已知知识.通过图像分类验证了本算法比其他迁移算法具有更好的分类精度.     

自适应耦合局部最优法及其在模型修正中的应用

耦合局部最优法作为一种新型的优化技术,既具有高效的搜索速度又具有全局搜索能力.然而,对于大规模优化问题,该方法容易陷入局部最优;另外,梯度信息在该项技术中起着重要作用,而对于复杂问题往往不能得到精确的梯度信息,从而使得该算法的全局搜索能力下降.本文分别从初始种群的确定、变步长搜索、自调节种群三方面对原算法进行了改进,提出了自适应耦合局部最优法,使之具备解决多变量复杂优化问题的能力.通过两个测试函数验证了改进算法比原有算法更易于得到全局最优解并保持较高的计算效率.最后,采用一个试验算例验证了自适应耦合局部最优法的有效性.     

Sequential and parallel schemes for adaptive 2-D parameter estimation with application to image estimation

In this paper, the non-causal quarter plane 2-D Recursive Least Squares (2D-RLS) algorithm for adaptive processing is developed. The complexity of this algorithm turns out to beO(L ⁶) per iteration, for anL ×L window. With the aim of reducing this complexity, the matrix gains appearing in the algorithm are replaced by scalar gains. This approach yields the Approximate 2-D Recursive Least Squares (A2D-RLS) algorithm, which is shown to have a complexity ofO(L ²). With the objective of reducing the computation time even further, a parallel scheme is developed for the A2D-RLS algorithm. Since the algorithm is inherently sequential, its parallelization involves some more approximations. The desired accuracy of the estimated parameters is shown to place an upper bound on the number of processors. The parallel scheme is suitable for implementation on shared memory as well as distributed memory machines. The algorithm is applied to the problem of image estimation. Simulation results giving speed-up, efficiency, and the accuracy of the estimated image are presented.     

Squeeze-and-breathe evolutionary Monte Carlo optimization with local search acceleration and its application to parameter fitting

Estimating parameters from data is a key stage of the modelling process, particularly in biological systems where many parameters need to be estimated from sparse and noisy datasets. Over the years, a variety of heuristics have been proposed to solve this complex optimization problem, with good results in some cases yet with limitations in the biological setting. In this work, we develop an algorithm for model parameter fitting that combines ideas from evolutionary algorithms, sequential Monte Carlo and direct search optimization. Our method performs well even when the order of magnitude and/or the range of the parameters is unknown. The method refines iteratively a sequence of parameter distributions through local optimization combined with partial resampling from a historical prior defined over the support of all previous iterations. We exemplify our method with biological models using both simulated and real experimental data and estimate the parameters efficiently even in the absence of a priori knowledge about the parameters.     

Tensor methods for parameter estimation and bifurcation analysis of stochastic reaction networks

Stochastic modelling of gene regulatory networks provides an indispensable tool for understanding how random events at the molecular level influence cellular functions. A common challenge of stochastic models is to calibrate a large number of model parameters against the experimental data. Another difficulty is to study how the behaviour of a stochastic model depends on its parameters, i.e. whether a change in model parameters can lead to a significant qualitative change in model behaviour (bifurcation). In this paper, tensor-structured parametric analysis (TPA) is developed to address these computational challenges. It is based on recently proposed low-parametric tensor-structured representations of classical matrices and vectors. This approach enables simultaneous computation of the model properties for all parameter values within a parameter space. The TPA is illustrated by studying the parameter estimation, robustness, sensitivity and bifurcation structure in stochastic models of biochemical networks. A Matlab implementation of the TPA is available at http://www.stobifan.org.     

Mobility model for heterogeneous wireless networks and its application in common radio resource management

User distribution and mobility behaviour vary based on environment types and characteristics. Heterogeneous wireless networks (HWNs) are deployed to utilise these characteristics and serve users with better quality. For efficient resource management in HWN environment, an understanding of multi-mode user mobility behaviour is paramount. Here, a multi-mode user mobility model is proposed in the context of wireless local area network (WLAN) coverage in the hotspot, overlaid on a macrocell of wireless wide area network (WWAN). An expression for microcell residence time of multi-mode users in HWNs is derived, based on the cell residence time in the constituting WLAN and WWAN. The boundary-crossing probabilities of moving into microcell, moving out of microcell and moving out of macrocell during a call for different types of hotspot topologies are also derived analytically. The numerical results obtained using the analytical expressions for boundary-crossing probability are validated by simulation results. The significance of the proposed mobility model is demonstrated through its application in common radio resource management (CRRM). Numerical results show that the mobility-based CRRM scheme exhibits a lower rate of unnecessary vertical handoffs than that achieved by the dasiaWLAN' if coveragedasia scheme that does not use mobility information for resource management.     

Neural networks for cost estimation: Simulations and pilot application

Neural networks in a multilayer perceptron architecture are able to classify data and approximate functions based on a set of sample data (curve fitting). These properties are used to investigate experimentally the applicability of neural networks for cost estimation in early phases of product design. Experiments are based on pilot cost data from a manufacturing company. In addition, artificially created simulative data are used for benchmarking. The cost estimation performance is compared to conventional methods, i.e. linear and non-linear parametric regression. Neural networks achieve lower deviations in their cost estimations. Beyond the use of standard neural architectures, simple modifications for a performance improvement are suggested and tested. Finally, a profile for situations where neural networks are appropriate is derived from the results.     

Dynamic electromagnetic torque model and parameter estimation for a deep-bar induction machine

A dynamic analytical electromagnetic torque model for a deep-bar induction machine is presented. The model is based on the space-vector theory of electrical machines. The parameters are estimated using the data provided by the numerical impulse response test performed within the two-dimensional time-stepping finite element analysis. Two different impulse tests are studied. The impulse excitation is applied to the stator voltage and rotor position angle, and the parameter estimates are compared. The applicability of the impulse test is based on the assumption of linear behaviour in the neighbourhood of an operation point. Hence, one of the main objectives is to study the validity of the impulse test. From the results of the angle impulse test, a frequency range of negative damping is detected. The phenomenon is also studied by measurements.     

Issues in optimal parameter estimation for the nonlinear Muskingum flood routing model

This study answers two questions raised in the parameter estimation optimization for the nonlinear Muskingum flood routing model. The first question is whether a new global optimum was still found after the existing global optimum had already been found. In order to fairly verify this question, a standard routing procedure for the nonlinear Muskingum model, which has not been clearly described previously, is proposed. Because the routing procedure was coded in a spreadsheet, any researcher can easily test it after downloading it. The second question is the reason why various approaches, such as Lagrange multiplier, Broyden–Fletcher–Goldfarb–Shanno (BFGS), genetic algorithm, harmony search and particle swarm optimization, have tackled only Wilson's data set as the parameter estimation optimization for the nonlinear Muskingum model, because Wilson's data have a unique structure which is differentiated from other data sets. This study also provides various data sets to compare.     

An algorithm for sub-optimal routeing in series-parallel queueing networks

The optimal routeing problem of maximizing system throughput in series-parallel networks with finite buffers is studied in this paper. The problem is extremely difficult to solve since closed form expressions are not easily constructed for throughput in finite networks. A piece-wise linear upper bound on the throughput of a tandem network is used to develop a throughput approximation in seriesparallel networks. Based on this approximation we are able to specify a suboptimal range for routeing probabilities at each junction in the network as a function of the arrival rate to this junction. We also specify a unique value for the routeing probability at each junction, independent of the arrival rate to that junction. We then construct an O(N) algorithm to analyse general series-parallel networks with more than one junction and specify the sub-optimal routeing probabilities at each junction.     

基于二项式R-D模型的多视点视频码率控制算法

针对目前尚未深入研究多视点视频编码(MVC)码率控制的状况,在分析现有视频码率控制中率失真模型的不足和多视点视频编码的特点的基础上,提出了一种基于二次率失真(R-D)模型的多视点视频编码码率控制算法。该算法的核心是先根据视差预测和运动预测的结构关系,将所有图像分成6种类型的编码帧,并改进二项式率失真模型,然后根据已编码信息进行视点间、帧层、基本单元层比特分配与码率控制。实验仿真结果表明,与目前采用固定量化参数的JVT的MVC相比,该算法能够有效地控制多视点视频编码的码率,同时保持高效的编码效率。     

Cooperative federated filtering approach for enhanced position estimation and sensor fault tolerance in ad-hoc vehicle networks

The present study addresses a solution to the position and orientation estimation problem of vehicles in ad-hoc vehicle networks using decentralised filtering. Specifically, a distributed filter operating in a cooperative federated structure for enhancing the estimation accuracy of vehicles state over unreliable wireless communication networks subject to uncertain and limited measurements is proposed. The filter relies on a variety of position measurements obtained from the on-board vehicle positioning system, from other cooperating vehicles in the vicinity, as well as from the immediate roadside environment via communication. Direct distance measurements between vehicles, furthermore, between vehicles and the stationary elements of the infrastructure can be used, if available, as highly accurate reference measurements. The study demonstrates how successfully the novel idea of cooperative federated filtering in addressing the demands of both fault tolerance and enhanced estimation accuracy in ad-hoc vehicle networks can be used.     

一种改进的UKF算法及其在深空自主光学导航中的应用

利用星敏感器以及光学导航相机,通过测量星光信息以及天体边缘信息,进行了自主光学导航方案的设计.通过观测量的转化改进了Unscented卡尔曼滤波方法(UKF)的具体实现形式,并将改进的方法与扩展的卡尔曼滤波方法(EKF)、UKF以及基于平方根分解的方法(SR-UKF)进行了比较,通过仿真对其算法的优越性进行了验证.仿真结果表明,这种基于观测量转换的UKF算法,不仅在计算量上有所减少,在精度上也有较大提高.     

一种改进的特征匹配算法在碎片云测量建模中的应用

为满足超高速撞击典型Whipple防护构型的损伤评估需求,利用图像处理技术对碎片云序列阴影图像进行深入研究.使用超高速序列激光阴影成像仪得到三组不同实验条件下碎片云发展过程的高清阴影图像,分别对每组最具代表性的2帧进行图像处理分析;根据碎片云图像特点以及碎片运动特性,提出了一种改进的碎片二次特征匹配算法,该方法包含碎片...     

汽车悬架磁流变减振器模型分析及半主动控制策略研究

用多项式模型作为磁流变减振器的力学模型,研究了传统"开-关"型半主动控制的阻尼力跳跃对汽车簧载质量振动加速度产生的不利影响,针对1/2汽车模型对具有磁流变减振器的汽车悬架简单模糊和自适应模糊神经半主动控制策略进行了数值仿真,基于数值仿真结果分析了各控制策略的效果.     

An equilibrium partitioning model for predicting response to host-guest complexation in electrospray ionization mass spectrometry

While electrospray ionization mass spectrometry (ESI-MS) has become a powerful technique for analyzing many types of host-guest complexation, questions remain as to just how accurately ion abundances generated by ESI reflect the true distribution of species at equilibrium in solution. To better understand this relationship, an equilibrium partitioning model was developed to explain the various interactions that dictate how much of a particular host-guest complex is transferred from solution into the gas phase in the ESI process. By evaluating the simultaneous equilibria of the complexation reaction and the partitioning of species between the surface and interior of the ESI droplets, one can estimate the ion abundances generated. The predictions of this new model were evaluated and experimentally confirmed through the analysis of the complexes of 18-crown-6 with alkali metal cations in an ESI quadrupole ion trap mass spectrometer, and it was determined that binding constants alone may not give accurate predictions about the observed ESI-MS response to different host-guest complexes.     

An iterative algorithm for solving inverse problems in structural dynamics

The pulse-spectrum technique (PST), an iterative numerical algorithm, is presented and extended to solve the inverse problems arising from the dynamic structural identification and structural design problems. A simple one-dimensional shear beam model is used to demonstrate the applicability of PST. Numerical simulations are carried out to test the feasibility and to study the general characteristics of this technique without the real measurement and design data. It is found that the PST is not only quite robust in providing accurate results but also an excellent numerical method according to the four practical criteria for evaluating numerical methods.     

前馈神经网络混合训练算法及其应用

介绍了单隐层前馈神经网络的混合训练算法(HFM)和正则化混合训练算法(RHFM),然后将该算法应用于UCI数据库上的实际回归例子中,并将其与BP、NNRW以及FM算法进行了比较.仿真实验表明,HFM算法的收敛速度优于其它几种算法,RHFM算法有较好的泛化性能,而NNRW算法在训练时间上占优,尽管如此,HFM算法在时间上还是大大优于BP算法.说明,混合训练算法是一种有效的算法.     

一种高效挖掘生物网络闭合频繁子图的算法

针对生物网络中频繁子图的挖掘问题,提出了一种基于FP-树结构的MaxFP算法.此算法以代谢路径作为研究对象,在适合于生物网络图简化模型的基础上,采用一种不产生候选集的改进FP-growth算法挖掘生物网络中的闭合频繁子图.此算法考虑了基于频繁项目集的算法应用于网络的缺陷,根据生物网络的特点对FP-growth算法进行了改进.实验证明,提出的MaxFP算法比基于Apriori的频繁模式挖掘算法运行速度快,不仅能挖掘出最大的频繁子图,且能找到更多具有生物意义的频繁子图.