用户浏览页面预测的多阶HMM模型融合

预测用户的浏览页面是WWW上的一个重要的研究方向.提出了一种通过模糊积分融合多阶HMM(Hidden Markov Model)预测结果的用户浏览页面预测模型.该算法首先拓展了经典融合预测算法的先验信息空间.其方法是通过对不同用户浏览模式分类,建立1阶多Morkov链模型并以其训练结果为权重指标,而后通过模糊积分理论融合1～N阶HMM预测的结果.性能测试实验结果表明该模型预测准确率优于已有的用户浏览页面预测的多HMM融合方法.该方法可在Web站点管理、电子商务以及网页预取等领域广泛应用.     

两步双向查找表预测的高光谱图像无损压缩

提出一种基于两步双向查找表预测的高光谱图像无损压缩算法。将谱段内预测和谱间预测有效地结合,去除了高光谱图像强的谱间相关性。根据高光谱图像特点,首先,在光谱线的第1谱段图像采用JPEG-LS中值预测器进行谱段内预测,其它谱段图像采用谱间预测。谱间预测采用两步双向预测算法,第1步预测,采用一种双向四阶预测器,利用该预测器得到参考预测值;第2步预测,采用一种8级查表(LUT)搜索预测算法,得出8个LUT预测值。然后,将参考预测值与其比较得出最终的预测值。最后,将预测差值进行熵编码。实验结果表明,本文算法的平均压缩比达到3.05bpp(bits per pixel),与传统高光谱图像无损压缩算法比较,平均压缩比提高了0.14～2.91bpp,有效提高了高光谱图像无损压缩比低的问题。     

混沌时间序列局域偏最小二乘回归多步预测模型

针对混沌序列局域一阶多步预测问题,提出了基于偏最小二乘回归的混沌时间序列局域直接多步预测模型,偏最小二乘用于混沌时序重构相空间中演化轨迹前后相点信息间的建模。该模型克服了以往一阶局域单步预测模型进行多步预测时存在的误差累积,而且能抑制重构相空间中多重共线性的影响,提高了预测精度。试验中使用交叉验证方法将偏最小二乘的提取成分数。通过对Chen’s混沌序列和Mackey-Glass混沌序列的多步预测试验,验证了该模型在混沌时序预测方面具有很好的效果。     

一类随机动态过程基于q阶树的多尺度建模方法

利用多尺度随机模型能建立处理问题有效并行算法的这一优势,提出一类随机动态过程基于一般q阶树的多尺度建模方法。首先,利用Markov过程的条件独立性给出一类过程基于q阶树的多尺度表示方法;其次,基于q阶树多尺度表示和具体实例推导出多尺度模型中的状态转移矩阵、扰动阵、初始状态和相应的协方差矩阵等的具体形式,为具有Markov统计特性的过程或信号建立起多尺度随机模型,这将为有效地解决多源同类信息和多源异类信息的数据融合等实际问题提供了理论基础;最后,给出一类Gauss-Markov过程基于三阶树和五阶树多尺度表示的计算机仿真结果,进一步验证建立模型的实用性和有效性。     

一种新的基于Markov链模型的用户行为异常检测方法

提出一种新的基于Markov链模型的用户行为异常检测方法。该方法利用一阶齐次Markov链对网络系统中合法用户的正常行为进行建模,将Markov链的状态同用户执行的shell命令序列联系在一起,并引入一个附加状态;在检测阶段,基于状态序列的出现概率对用户当前行为的异常程度进行分析,并根据Markov链状态的实际含义和用户行为的特点, 采用了较为特殊的判决准则。与Lane T提出的基于隐Markov模型的检测方法相比,该方法的计算复杂度较低,更适用于在线检测。而同基于实例学习的检测方法相比,该方法则在检测准确率方面具有较大优势。文中提出的方法已在实际入侵检测系统中得到应用,并表现出良好的检测性能。     

基于加权一阶局域法的负荷区间预测研究

文中将加权一阶局域法应用到负荷区间预测法中,综合得到复杂电力系统在单步负荷区间预测与多步负荷区间预测中的基本模型,并说明加权一阶局域法的基本过程,从理论上分析加权一阶局域法的多步负荷区间预测曲线相交叉的原因。采用该方法进行基于Lorenz方程和PJM实际电力负荷数据的区间预测,结果证明受曲线相点交叉的影响导致在多步负荷区间预测上部分点的数值只有一个公共值而无法构成区间,因此基于加权一阶局域法的多步负荷区间预测存在相对明显的应用局限性。     

改进型可过滤混合分支预测器

针对TAGE混合预测器T0表内容存在大量混叠以及TAGE混合预测器对历史相关性较低的指令预测准确率低的问题,文章提出了一种新的解决方案。该方案采用YAGS作为TAGE的基础预测器,并在TAGE前后各加一个过滤器,前端过滤器Filter1专用于过滤特定循环分支指令,后端过滤器Filter2专用于过滤和全局历史相关性低的分支指令。在CBP-2模拟环境下进行实验,同时测试改进后设计的预测器与现有的TAGE混合预测器,在256k硬件资源配置下,改进后的设计预测准确率指标为3.972MPKI,优于现有TAGE的4.411MPKI。由此可见改进后的设计具有一定的参考意义。     

基于EEMD技术在电力信息安全中的多步时间序列预测方法

针对用户访问轨迹的数据特征,提出一种基于EEMD技术的多步时间序列预测模型。该模型利用了集合经验模态分解EEMD结合极限学习机ELM模型,混合人工鱼群MAFA优化的方式,克服了算法中存在过拟合和多步时间序列预测的策略限制问题。通过该模型,实现了对访问轨迹时间序列多步预测,结合安全范围包络线,进而提前发现是否存在入侵行为。验证结果表明,优化后的EEMD-ELM模型比传统时间序列预测方法的迭代速率与精度得到了极大提高,泛化能力增强,说明了该方法的有效性、可行性。     

基于多Markov链预测模型的Web缓存替换算法

为了提高web缓存的性能,提出了一种基于多Markov链预测模型的Web缓存替换算法PGDSF-AI.首先将Web中具有不同浏览特征的用户分为多类,为每一类用户建立类Markov链,进一步建立多Markov链预测模型.然后利用该模型对当前的用户请求预测,进而组成预测对象集.当缓存空间不足时,选取键值最小且不在预测对象集中的对象替换.通过估算对象的平均间隔时间,避免缓存大量保留长时间没有访问的对象.实验结果表明,提出的算法有较好的性能.     

时变信道中自适应VR-HARQ研究

为了更好地解决时变信道中可靠性与吞吐率这一对矛盾,本文提出一种基于线性预测的自适应冗余可变混合ARQ(VR-HARQ)方案.在该方案中,由于时变信道的慢衰落特性,时变信道被等效为有限状态的Markov过程,每个Markov状态对应信道的不同信噪比,同时,文章详细地描述了信道吞吐率与误码率之间的数学关系,推导了每个Markov状态的最佳编码方案,然后系统采用自适应线性预测算法,根据当前个时段的信道状态,估计下一个时段信道的信噪比以及其所对应的Markov状态,最后根据吞吐率最大原则选择合适的纠错编码方案.仿真结果表明:自适应线性预测VR-HARQ方案的性能明显优于传统VR-HARQ方案.     

Throughput analysis method for hybrid ARQ schemes over burst errorchannels

A definition of a burst error channel using a Markov model was presented by T. Sato et al. in a previous paper (1991). A throughput analysis method of hybrid automatic repeat request (ARQ) under the burst error channel using the three-state Markov model is described. The hybrid ARQ is studied for the random and burst error correction codes as the forward error correction (FEC) code, and multiframe rejection (MREJ) as the ARQ. The throughput efficiency is obtained with both an infinite buffer memory and a finite buffer memory. The applicable range of the burst error channel is clarified for the hybrid ARQ using random and burst error correction codes     

Hybrid approach to speech recognition using hidden Markov modelsand Markov chains

The paper presents a hybrid of a hidden Markov model and a Markov chain model for speech recognition. In this hybrid, the hidden Markov model is concerned with the time-varying property of spectral features, while the Markov chain accounts for the interdependence of spectral features. The log-likelihood scores of the two models, with respect to a given utterance, are combined by a postprocessor to yield a combined log-likelihood score for word classification. Experiments on speaker-independent and multispeaker isolated English alphabet recognition show that the hybrid outperformed both the hidden Markov model and the Markov chain model in terms of recognition     

Linear-Complexity Models for Wireless MAC-to-MAC Channels

Wireless local area networks suffer from frequent bit-errors that result in Medium Access Control (MAC) layer packet drops. Bandwidth and media quality constraints of real-time applications necessitate analysis and modeling at the “MAC-to-MAC wireless channel”. In this paper, we propose and evaluate stochastic models for the 802.11b MAC-to-MAC bit-error process. We propose an Entropy Normalized Kullback-Leibler (ENK) measure to accurately evaluate the performance of the models. We employ this measure to demonstrate that the traditional full-state Markov chains of order-10 and order-9 are required for accurate representation of the channel at 2 and 5.5 Mbps, respectively. However, the complexity of this modeling paradigm increases exponentially with respect to the order. For many real-time and non-real-time applications, which require (or could benefit significantly from) accurate modeling, the high complexity of full-state high-order Markov models makes them impractical or virtually ineffective. Thus, we propose two new linear-complexity models, which we refer to as the short-term energy model (SEM) and the zero-crossing model (ZCM). These models, which constitute the most important contribution of this paper, constrain the complexity to increase linearly with the model order. We illustrate that the linear-complexity models, while yielding orders of magnitude reduction in complexity, provide a performance comparable to that of the exponential complexity full-state models. Within the linear-complexity context, we illustrate that the zero-crossing model perform better than its short-term energy counterpart. Finally, for varying window sizes and due to its low complexity, we show that the zero-crossing model can be adapted in real-time. Such an adaptive model provides accurate channel modeling and characterization for rate adaptive applications.Syed Ali Khayam received the B.S. degree in computer systems engineering from National University of Sciences and Technology (NUST) Pakistan in 1999. He secured the Pakistan Higher Education Commission M.S./Ph.D. scholarship to pursue post graduate studies at Michigan State University (MSU). He completed his M.S. in Electrical Engineering from MSU in 2003. He is currently a Ph.D. candidate in the Department of Electrical and Computer Engineering at MSU. He also worked at Communications Enabling Technologies where he led a design team which realized various modules of a Voice-over-IP system-on-chip. His research interests include statistical analysis and modeling of computer (and in particular wireless) networks, network security, cross-layer protocol design, real-time multimedia communications over IP-based networks, and VLSI chip design.Hayder Radha received the B.S. degree (with honors) from Michigan State University (MSU) in 1984, the M.S. degree from Purdue University in 1986, and the Ph.M. and Ph.D. degrees from Columbia University in 1991 and 1993 (all in electrical engineering). He joined MSU in 2000 as Associate Professor in the Department of Electrical and Computer Engineering. Between 1996 and 2000, Dr. Radha worked at Philips Research USA where he initiated the Internet Video project and led a team of researchers working on scalable video coding and streaming algorithms. Dr. Radha is a Philips Research Fellow. Prior to working at Philips, Hayder Radha was a Distinguished Member of Technical Staff at Bell Labs where he worked between 1986 and 1996 in the areas of digital communications, signal/image processing, and broadband multimedia. His research interests include image and video coding, wireless technology, multimedia communications and networking. He has more than 20 patents in these areas. Dr. Radha served as Co-Chair and Editor of the ATM and LAN Video Coding Experts Group of the ITU-T in 1994–1996.     

MEI系数的快速算法

不变性测试方程法已被证明是解决电磁问题的一种有效方法。目前电大尺寸问题中ＭＥＩ系数的计算已成为一个瓶颈。提出了一个快速算法用于加速ＭＥＩ系数的计算,它使用快速多极子方法计算测试子的散射场,使得ＭＥＩ系数的计算速度从Ｏ（Ｎ＾２）变为Ｏ（Ｎ＾１．５Ｌｏｇ２Ｎ）。     

Finite-memory universal prediction of individual sequences

The problem of predicting the next outcome of an individual binary sequence under the constraint that the universal predictor has a finite memory, is explored. In this analysis, the finite-memory universal predictors are either deterministic or random time-invariant finite-state (FS) machines with K states (K-state machines). The paper provides bounds on the asymptotic achievable regret of these constrained universal predictors as a function of K, the number of their states, for long enough sequences. The specific results are as follows. When the universal predictors are deterministic machines, the comparison class consists of constant predictors, and prediction is with respect to the 0-1 loss function (Hamming distance), we get tight bounds indicating that the optimal asymptotic regret is 1/(2K). In that case of K-state deterministic universal predictors, the constant predictors comparison class, but prediction is with respect to the self-information (code length) and the square-error loss functions, we show an upper bound on the regret (coding redundancy) of O(K/sup -2/3/) and a lower bound of /spl Theta/(K/sup -4/5/). For these loss functions, if the predictor is allowed to be a random K-state machine, i.e., a machine with random state transitions, we get a lower bound of /spl Theta/(1/K) on the regret, with a matching upper bound of O(1/K) for the square-error loss, and an upper bound of O(logK/K) Throughout the paper for the self-information loss. In addition, we provide results for all these loss functions in the case where the comparison class consists of all predictors that are order-L Markov machines.     

针对时间序列多步预测的聚类隐马尔科夫模型

时间序列的预测在现今社会各个领域中有着广泛的应用.本文针对时间序列趋势预测中的多步预测问题,提出了基于聚类的隐马尔科夫模型,利用隐马尔科夫模型中的隐状态来表示产生时间序列数据时的系统内部状态,实现对多步时间序列的预测.针对时间序列聚类中的距离计算问题,提出结合时间序列时间性和相似性的聚类算法,并给出了迭代精化基于聚类的隐马尔科夫模型的方法.实验表明,本文提出的方法在时间序列多步预测中精度较高.     

A mixed MAP/MLSE receiver for convolutional coded signalstransmitted over a fading channel

This paper addresses the problem of estimating a rapidly fading convolutionally coded signal such as might be found in a wireless telephony or data network. We model both the channel gain and the convolutionally coded signal as Markov processes and, thus, the noisy received signal as a hidden Markov process (HMP). Two now-classical methods for estimating finite-state hidden Markov processes are the Viterbi (1967) algorithm and the a posteriori probability (APP) filter. A hybrid recursive estimation procedure is derived whereby one hidden process (the encoder state in our application) is estimated using a Viterbi-type (i.e., sequence based) cost and the other (the fading process) using an APP-based cost such as maximum a posteriori probability. The paper presents the new algorithm as applied specifically to this problem but also formulates the problem in a more general setting. The algorithm is derived in this general setting using reference probability methods. Using simulations, performance of the optimal scheme is compared with a number of suboptimal techniques-decision-directed Kalman and HMP predictors and Kalman filter and HMP filter per-survivor processing techniques     

Modeling heterogeneous network traffic in wavelet domain

Heterogeneous network traffic possesses diverse statistical properties which include complex temporal correlation and non-Gaussian distributions. A challenge to modeling heterogeneous traffic is to develop a traffic model which can accurately characterize these statistical properties, which is computationally efficient, and which is feasible for analysis. This work develops wavelet traffic models for tackling these issues. We model the wavelet coefficients rather than the original traffic. Our approach is motivated by a discovery that although heterogeneous network traffic has the complicated short- and long-range temporal dependence, the corresponding wavelet coefficients are all “short-range” dependent. Therefore, a simple wavelet model may be able to accurately characterize complex network traffic. We first investigate what short-range dependence is important among the wavelet coefficients. We then develop the simplest wavelet model, i.e., the independent wavelet model for Gaussian traffic. We define and evaluate the (average) autocorrelation function and the buffer loss probability of the independent wavelet model for fractional Gaussian noise (FGN) traffic. This assesses the performance of the independent wavelet model, and the use of which for analysis. We also develop (low-order) Markov wavelet models to capture additional dependence among the wavelet coefficients. We show that an independent wavelet model is sufficiently accurate, and a Markov wavelet model only improves the performance marginally. We further extend the wavelet models to non-Gaussian traffic through developing a novel time-scale shaping algorithm. The algorithm is tested using real network traffic and shown to outperform FARIMA in both efficiency and accuracy. Specifically, the wavelet models are parsimonious, and have a computational complexity O(N) in developing a model from a training sequence of length N, and O(M) in generating a synthetic traffic trace of length M