发现和学习不可复位动态系统的预测状表示的一种新算法

提出了一种发现和学习不可复位动态系统的预测状态表示的新算法.在证明系统的任意landmark均可作为系统的初始状态的基础上,利用发现的landmark确定系统在任意时间步所处的经历,然后采用蒙特卡罗方法估计任意经历下任意检验发生的概率,解决了在不可复位动态系统中,经历下检验发生的概率难以获取问题,进而发现和学习不可复位动态系统的预测状态表示.实验结果表明,本文算法获得的系统的预测状态表示在预测精度上明显优于suffix-history算法,验证了所提算法的有效性.     

基于确定有穷自动机描述的仿人智能控制记忆

模仿熟练操作者通过记忆多步连续调控系统的方式,将记忆用上下文无关文法表示.根据控制经验和滚动预测优化建立特征状态转换表和构造不确定有穷自动机,给出了转换确定有穷自动机的算法.在任意特征状态下,通过一系列的DFA状态转换函数的复合运算,得到使系统稳定下来的控制模态序列.     

快变信道环境下基于支持向量回归的频谱预测

频谱预测是一种通过分析历史频谱数据获得频谱使用规律,从而预测未来频谱使用状态的技术。为了实现快变信道（本文指信道占用状态快速变化）环境下频谱状态的可靠预测,提出了一种基于支持向量回归的频谱预测算法。比较了在不同训练样本数时,该算法与一个典型的BP神经网络频谱预测算法的性能差异,结果表明所提算法在小样本学习时,预测效果更为理想。并在此基础上,加入正确检测概率和虚警概率,验证了当频谱检测不理想条件下,支持向量回归算法预测的可行性。      

虚拟化云无线接入网络下基于在线学习的网络切片虚拟资源分配算法

针对现有研究中缺乏云无线接入网络(C-RAN)场景下对网络切片高效的动态资源分配方案的问题,该文提出一种虚拟化C-RAN网络下的网络切片虚拟资源分配算法。首先基于受限马尔可夫决策过程(CMDP)理论建立了一个虚拟化C-RAN场景下的随机优化模型,该模型以最大化平均切片和速率为目标,同时受限于各切片平均时延约束以及网络平均回传链路带宽消耗约束。其次,为了克服CMDP优化问题中难以准确掌握系统状态转移概率的问题,引入决策后状态(PDS)的概念,将其作为一种“中间状态”描述系统在已知动态发生后,但在未知动态发生前所处的状态,其包含了所有与系统状态转移有关的已知信息。最后,提出一种基于在线学习的网络切片虚拟资源分配算法,其在每个离散的资源调度时隙内会根据当前系统状态为每个网络切片分配合适的资源块数量以及缓存资源。仿真结果表明,该算法能有效地满足各切片的服务质量(QoS)需求,降低网络回传链路带宽消耗的压力并同时提升系统吞吐量。     

基于动态马尔科夫模型的入侵检测技术研究

本文提出了基于动态马尔科夫模型的入侵检测方法.首先提取特权进程的行为特征,并在此基础上动态构造Markov模型.由动态Markov模型产生的状态序列计算状态概率,根据状态序列概率来评价进程行为的异常情况.利用Markov模型的动态构造充分提取特权进程的局部行为特征的相互关系,因此可以在训练数据集有限的条件下使模型更精确、检测能力大大加强.实验表明该算法准确率高、实时性强、占用系统资源少.本文所提方法算法简单、预测准确,适合于进行实时检测.     

基于车车通信的车辆防碰撞算法

为了提前预测车辆碰撞事故和降低碰撞事故发生概率,提出了基于车车通信的车辆防碰撞算法（VACA）。VACA根据自身和周围邻居车辆的北斗位置坐标,在水平和垂直两个方向上分别采用Kalman预测算法预测下一时刻的车辆行驶状态,采用数据预判断方法直接确定一定发生碰撞或不发生碰撞的情况。当不能直接判断时,则计算加速度可变下的安全距离,建立和求解碰撞状态预测模型,获得碰撞预测时间最小值。如果该最小值小于阈值,则发出报警信号提醒驾驶员。仿真结果表明,与基于车辆间通信的智能防碰撞报警模型（IVCWM）和车辆提前预警的防碰撞算法（ECWA）比较,VACA可提前发现碰撞情况,而且更准确,有一定的应用价值。     

动态网络表示学习研究进展

网络表示学习旨在将网络信息表示为低维稠密的实数向量,解决链接预测、异常检测、推荐系统等任务.近年来,网络表示学习研究取得重大进展,但研究多基于静态网络,而真实世界构成的网络是动态变化的,对动态网络分析的需求日益增加.本文总结了当前动态网络表示学习的方法与研究进展,首先提出网络表示学习的动机,阐述动态网络以及表示学习的发展历史与理论基础;接着,系统概述了大量动态网络嵌入方法,包括基于矩阵分解的动态图嵌入、基于随机游走的动态图嵌入、基于深度学习的动态图嵌入和基于重构概率的动态图嵌入,并分析与比较,给出动态网络表示学习的应用场景;最后,总结未来网络表示学习的研究方向.只有考虑网络的动态性,才能真实反映现实网络的演化,使网络表示学习更具价值.     

采用HDPHMM符号化器的语音查询样例检测方法

提出一种基于层级狄利克雷过程隐马尔科夫模型（HDPHMM）符号化器的无监督语音查询样例检测（QbE-STD）方法。该方法首先应用一个双状态层隐马尔科夫模型,其中顶层状态用于表示所发现的声学单元,底层状态用于建模顶层状态的发射概率,通过对顶层状态假设一个层级狄利克雷过程先验,获得非参贝叶斯模型HDPHMM。使用无标注语音数据对该模型进行训练,然后对测试语音和查询样例输出后验概率特征矢量,使用非负矩阵分解算法对后验概率进行优化得到新的特征,然后在此基础上,应用修正分段动态时间规整算法进行检索,构成QbE-STD系统。实验结果表明,相比于基于高斯混合模型符号化器的基线系统,本文所提出的方法性能更优,检索精度得到显著提升。      

嵌入式系统动态电源管理预测算法研究

针对目前在嵌入式系统中动态电源管理普遍采用超时算法节能效率低的缺点,采用预测的方法进行动态电源管理.提出一个动态电源管理预测算法,该算法通过利用设备空闲状态的历史信息对未来的空闲时间进行预测,预测结果作为动态电源管理的依据.实验结果说明该算法对工作状态平稳的系统空闲预测效果比较理想,适用于动态电源管理.     

基于ZigBee技术的RFID空间定位系统

通过ZigBee mote与RFID reader结合的方式应用随机数定位算法展示了一种低能耗的基于Zigbee技术的RFID空间定位系统,使得对佩带了Zigbee mote的人可以实时进行定位.在本系统中,通过使用基于取样的表示方法,定位算法能够表示任意分布.通过将系统实现的算法与算法原型比较,可以发现在Non-Line-Of-Sight(NLOS)场景下,本算法的定位错误(positioning errors)有明显改进.     

Nonlinear estimation with quantized measurements--PCM, predictive quantization, and data compression

Statistics conditioned on quantized measurements are considered in the general case. These results are specialized to Gaussian parameters and then extended to discrete-time linear systems. The conditional mean of the system's state vector may be found by passing the conditional mean of the measurement history through the Kalman filter that would be used had the measurements been linear. Repetitive use of Bayes' rule is not required. Because the implementation of this result requires lengthy numerical quadrature, two approximations are considered: the first is a power-series expansion of the probablity-density function; the second is a discrete-time version of a previously proposed algorithm that assumes the conditional distribution is normal. Both algorithms may be used with any memory length on stationary or nonstationary data. The two algorithms are applied to the noiseless-channel versions of the PCM, predictive quantization, and predictive-comparison data compression systems; ensemble-average performance estimates of the nonlinear filters are derived. Simulation results show that the performance estimates are quite accurate for most of the cases tested.     

Toward a deeper understanding of the availability of series-systems without aging during repairs

It is shown in detail that the steady-state availability of a series-system with no aging of components during the repair of another one can be interpreted as a conditional probability of the well-known s-independent case conditioned on allowing for at most one failed component at any time. Trivially, this conditional probability is larger than the standard probability, viz. the product of all components' probabilities. Yet it can be shown that also without this new interpretation of the formula for the availability of a series-system without aging during repairs it is exceptionally easy to prove the fact that by no aging during repairs, the availability of series-systems increases; a most plausible result.     

Extensions of the SMC-PHD filters for jump Markov systems

The probability hypothesis density (PHD) filter is a promising algorithm for multitarget tracking, which can be extended for jump Markov systems (JMS). Since the existing multiple model sequential Monte Carlo PHD (MM SMC-PHD) filter is not interacting, two extensions of the SMC-PHD filters are developed in this paper. The interacting multiple-model (IMM) SMC-PHD filter approximates the model conditional PHD of target states by particles, and performs the interaction by resampling without any a priori assumption of the noise. The IMM Rao-Blackwellized particle (RBP) PHD filter uses the idea of Rao-Blackwellized to further enhance the performance of target state estimation for JMS with mixed linear/nonlinear state space models. The simulation results show that the proposed algorithms have better performances than the existing MM SMC-PHD filter in terms of state filtering and target number estimation.     

Prony and Polynomial Approximations for Evaluation of the Average Probability of Error Over Slow-Fading Channels

A novel and simple semianalytical method for evaluating the average probability of transmission error for digital communication systems that operate over slow-fading channels is presented. The proposed method applies a sum of exponentials fit known as the Prony approximation to the conditional probability of error. Hence, knowledge of the moment-generating function of the instantaneous signal-to-noise ratio (SNR) at the detector input can be used to obtain the average probability of error. Numerical results show that knowledge of the conditional probability of error at only a small number of points and the sum of only two exponentials are sufficient to achieve very high accuracy; the relative approximation error of the exact average probability of error is less than 6% in most of the cases considered. Furthermore, a piecewise polynomial approximation of the conditional probability of error is investigated as an alternative to the sum of exponentials fit. In this case, knowledge of the partial moments of the instantaneous SNR at the detector input can be used to obtain the average probability of error. Numerical results indicate that, to achieve good accuracy, the method based on the polynomial approximation requires that the product of the polynomial degree and the number of approximation subintervals be larger than 10.      

Approximation of dynamical time-variant systems by continuous-time recurrent neural networks

This paper studies the approximation ability of continuous-time recurrent neural networks to dynamical time-variant systems. It proves that any finite time trajectory of a given dynamical time-variant system can be approximated by the internal state of a continuous-time recurrent neural network. Given several special forms of dynamical time-variant systems or trajectories, this paper shows that they can all be approximately realized by the internal state of a simple recurrent neural network.     

Analysis of conditional MTTF of fault-tolerant systems

Mean time to failure (MTTF) is one of the most frequently used dependability measures in practice. By convention, MTTF is the expected time for a system to reach any one of the failure states. For some systems, however, the mean time to absorb to a subset of the failure states is of interest. Therefore, the concept of conditional MTTF may well be useful. In this paper, we formalize the definition of conditional MTTF and cumulative conditional MTTF with an efficient computation method in a finite state space Markov model. Analysis of a fault-tolerant disk array system and a fault-tolerant software structure are given to illustrate application of the conditional MTTF.     

A Posteriori Probability Distances Between Finite-Alphabet Hidden Markov Models

In this correspondence, we consider a probability distance problem for a class of hidden Markov models (HMMs). The notion of conditional relative entropy between conditional probability measures is introduced as an a posteriori probability distance which can be used to measure the discrepancy between hidden Markov models when a realized observation sequence is observed. Using a measure change technique, we derive a representation for conditional relative entropy in terms of the parameters of the HMMs and conditional expectations given measurements. With this representation, we show that this distance can be calculated using an information state approach     

A lower bound on authentication after having observed a sequence of messages

In this paper we study authentication systems and consider the following scenario: Each encoding rule is used for the transmission of a sequence of i messages. We prove a lower bound on the probability that a spoofer observing i messages succeeds in generating an authentic message without knowing the encoding rule used. This bound is based on the conditional entropy of the encoding rules when a sequence of messages is known. Authentication systems which meet the bound are investigated and compared with systems that are l-fold secure against spoofing introduced by Massey [8]. We also give a bound for the probability of success if the opponent can choose how many messages he observes before trying to cheat.     

Nonlinear Bayesian Filtering Using the Unscented Linear Fractional Transformation Model

For nonlinear state space model involving random variables with arbitrary probability distributions, the state estimation given a sequence of observations is based on an appropriate criterion such as the minimum mean square error (MMSE). This leads to linear approximation in the state space of the extended Kalman filter (EKF) and the unscented Kalman filter (UKF), which work reasonably well only for mildly nonlinear systems. We propose a Bayesian filtering technique based on the MMSE criterion in the framework of the virtual linear fractional transformation (LFT) model, which is characterized by a linear part and a simple nonlinear structure in the feedback loop. LFT is an exact representation for any differentiable nonlinear mapping, so the virtual LFT model is amenable to a wide range of nonlinear systems. Simulation results demonstrate that the proposed filtering technique gives better approximation and tracking performance than standard methods like the UKF. Furthermore, for highly nonlinear systems where UKF diverges, the LFT model estimates the conditional mean with reasonable accuracy.      

Posterior Cramer-Rao bounds for discrete-time nonlinear filtering

A mean-square error lower bound for the discrete-time nonlinear filtering problem is derived based on the van Trees (1968) (posterior) version of the Cramer-Rao inequality. This lower bound is applicable to multidimensional nonlinear, possibly non-Gaussian, dynamical systems and is more general than the previous bounds in the literature. The case of singular conditional distribution of the one-step-ahead state vector given the present state is considered. The bound is evaluated for three important examples: the recursive estimation of slowly varying parameters of an autoregressive process, tracking a slowly varying frequency of a single cisoid in noise, and tracking parameters of a sinusoidal frequency with sinusoidal phase modulation