A Hybrid Method Using HAVOK Analysis and Machine Learning for Predicting Chaotic Time Series

The prediction of chaotic time series systems has remained a challenging problem in recent decades. A hybrid method using Hankel Alternative View Of Koopman (HAVOK) analysis and machine learning (HAVOK-ML) is developed to predict chaotic time series. HAVOK-ML simulates the time series by reconstructing a closed linear model so as to achieve the purpose of prediction. It decomposes chaotic dynamics into intermittently forced linear systems by HAVOK analysis and estimates the external intermittently forcing term using machine learning. The prediction performance evaluations confirm that the proposed method has superior forecasting skills compared with existing prediction methods.     

A new local linear prediction model for chaotic time series

A new local linear prediction model is proposed to predict chaotic time series in this Letter. We propose that the parameters—the embedding dimension and the time delay of the local linear prediction model—can take values which are different to those of the state space reconstruction in the procedure of finding the nearest neighbor points. We propose a criterion based on prediction power to determine the optimal parameters of the new local linear prediction model. Simulation results show that the new local linear prediction model can effectively predict chaotic time series and the prediction performance of the new local linear prediction model is superior to that of the local linear prediction.     

The improved local linear prediction of chaotic time series

Based on the Bayesian information criterion, this paper proposes the improved local linear prediction method to predict chaotic time series. This method uses spatial correlation and temporal correlation simultaneously. Simulation results show that the improved local linear prediction method can effectively make multi-step and one-step prediction of chaotic time series and the multi-step prediction performance and one-step prediction accuracy of the improved local linear prediction method are superior to those of the traditional local linear prediction method.     

Analyses of Optimal Embedding Dimension and Delay for Local Linear Prediction Model

In the reconstructed phase space, a novel local linear prediction model is proposed to predict chaotic time series. The parameters of the proposed model take the values that are different from those of the phase space reconstruction. We propose a criterion based on prediction error to determine the optimal parameters of the proposed model. The simulation results show that the proposed model can effectively make one-step and multistep prediction for chaotic time series, and the one-step and multi-step prediction accuracy of the proposed model is superior to that of the traditional local linear prediction.     

基于SARIMA-BP神经网络组合方法的MODIS叶面积指数时间序列建模与预测

植被叶面积指数(LAI)时间序列的建模及预测是陆面过程模型和遥感数据同化方法的重要组成部分。MODIS数据产品MOD15A2是目前应用最为广泛的LAI数据源之一,然而MODIS LAI时间序列产品包含了一些低质量的数据,例如由于云层、气溶胶等的影响,该产品在时间和空间上缺乏连续性。MODIS LAI时间序列包含线性部分和外在干扰产生的非线性部分,单一的线性方法或非线性方法都不能对其精确建模和预测。首先利用Savitzky-Golay(SG)滤波和线性插值平滑受到干扰的LAI时间序列,然后采用季节自回归积分滑动平均(SARIMA)方法、BP神经网络方法及二者的组合方法(SARIMA-BP)对MODIS LAI时间序列进行建模及预测。在SARIMA-BP神经网络组合方法中,各自在线性与非线性建模的优势得以充分发挥,其中SARIMA方法用于建模及预测LAI时间序列中的线性部分,BP神经网络方法用于对非线性残差部分进行建模及预测。实验结果显示：SG滤波和线性插值后的LAI时间序列比原LAI时间序列更平滑;SARIMA-BP神经网络组合方法的决定系数为0.981,比SARIMA和BP神经网络的0.941和0.884更接近于1;SARIMA-BP神经网络组合方法的预测值同观测值之间的相关系数为0.991,高于SARIMA(0.971)和BP神经网络(0.942)的相关系数。由此得出结论：SARIMA-BP神经网络组合方法对MODIS LAI时间序列具有更好的适应性,其建模和预测准确性高于SARIMA方法或BP神经网络方法。     

Rotational Diffusion in Prolate Symmetric top Molecules

A comparative analysis of theoretical and experimental rotational diffusion constants was performed for a series of liquid phase prolate symmetric top molecules. It was observed that the tumbling diffusion constants, D⊥, predicted by the extended Hynes-Kapral Weinberg (HKW) model were invariably in closer agreement with experiment than were values determined by the Hu-Zwanzig (HZ) slip model of molecular reorientation. The HKW theory also yielded superior estimates of D‖ for two -CCI³ rotors. However, the experimental spinning diffusion coefficients for methyl (-CD³) rotors were much closer to the slip (Free Rotor) limit.     

Don't bleach chaotic data

A common first step in time series signal analysis involves digitally filtering the data to remove linear correlations. The residual data is spectrally white (it is "bleached"), but in principle retains the nonlinear structure of the original time series. It is well known that simple linear autocorrelation can give rise to spurious results in algorithms for estimating nonlinear invariants, such as fractal dimension and Lyapunov exponents. In theory, bleached data avoids these pitfalls. But in practice, bleaching obscures the underlying deterministic structure of a low-dimensional chaotic process. This appears to be a property of the chaos itself, since nonchaotic data are not similarly affected. The adverse effects of bleaching are demonstrated in a series of numerical experiments on known chaotic data. Some theoretical aspects are also discussed.     

Resummation of QED perturbation series by sequence transformations and the prediction of perturbative coefficients

We propose a method for the resummation of divergent perturbative expansions in quantum electrodynamics and related field theories. The method is based on a nonlinear sequence transformation and uses as input data only the numerical values of a finite number of perturbative coefficients. The results obtained in this way are for alternating series superior to those obtained using Pade approximants. The nonlinear sequence transformation fulfills an accuracy-through-order relation and can be used to predict perturbative coefficients. In many cases, these predictions are closer to available analytic results than predictions obtained using the Pade method.     

Multivariate Time Series Imputation: An Approach Based on Dictionary Learning

The problem addressed by dictionary learning (DL) is the representation of data as a sparse linear combination of columns of a matrix called dictionary. Both the dictionary and the sparse representations are learned from the data. We show how DL can be employed in the imputation of multivariate time series. We use a structured dictionary, which is comprised of one block for each time series and a common block for all the time series. The size of each block and the sparsity level of the representation are selected by using information theoretic criteria. The objective function used in learning is designed to minimize either the sum of the squared errors or the sum of the magnitudes of the errors. We propose dimensionality reduction techniques for the case of high-dimensional time series. For demonstrating how the new algorithms can be used in practical applications, we conduct a large set of experiments on five real-life data sets. The missing data (MD) are simulated according to various scenarios where both the percentage of MD and the length of the sequences of MD are considered. This allows us to identify the situations in which the novel DL-based methods are superior to the existing methods.     

Fluctuations of the quark densities in nuclei

We study the static scalar susceptibility of the nuclear medium, i.e., the change of the quark condensate for a small modification of the quark mass. In the linear sigma model it is linked to the in-medium sigma propagator and its magnitude increases due to the mixing with the softer modes of the nucleon-hole excitations. We show that the pseudoscalar susceptibility, which is large in the vacuum, owing to the smallness of the pion mass, follows the density evolution of the quark condensate and thus decreases. At normal nuclear matter density the two susceptibilities become much closer, a partial chiral symmetry restoration effect as they become equal when the full restoration is achieved. Received: 20 July 2002 / Accepted: 14 September 2002 / Published online: 21 January 2003 RID="a" ID="a"e-mail: chanfray@ipnl.in2p3.fr Communicated by A. Molinari     

线性非高斯序列的替代数据检验新方法

替代数据法作为检验时间序列非线性和混沌的统计方法获得了广泛的应用.由于原替代数据法的零假设为线性高斯过程,可能把线性非高斯过程,特别是非最小相位过程误判为非线性.为了解决这一问题,提出并详细推导了基于功率谱等价的非最小相位序列求逆方法;结合基于高阶累积量的非最小相位自回归滑动平均模型辨识方法,提出了检验序列是否为线性非高斯过程的替代数据生成新算法.仿真算例表明,上述方法成功地克服了原替代数据法的不足. 关键词： 替代数据 非线性检验 非最小相位 功率谱等价     

Stock Index Prediction Based on Time Series Decomposition and Hybrid Model

The stock index is an important indicator to measure stock market fluctuation, with a guiding role for investors’ decision-making, thus being the object of much research. However, the stock market is affected by uncertainty and volatility, making accurate prediction a challenging task. We propose a new stock index forecasting model based on time series decomposition and a hybrid model. Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) decomposes the stock index into a series of Intrinsic Mode Functions (IMFs) with different feature scales and trend term. The Augmented Dickey Fuller (ADF) method judges the stability of each IMFs and trend term. The Autoregressive Moving Average (ARMA) model is used on stationary time series, and a Long Short-Term Memory (LSTM) model extracts abstract features of unstable time series. The predicted results of each time sequence are reconstructed to obtain the final predicted value. Experiments are conducted on four stock index time series, and the results show that the prediction of the proposed model is closer to the real value than that of seven reference models, and has a good quantitative investment reference value.     

基于相空间邻域的混沌时间序列自适应预测滤波器(Ⅰ)线性自适应滤波

提出了一种基于相空间邻域的线性自适应滤波算法，将时间域变换到高维重构矢量空间，使 得不可能实现的线性预测变成了可能.实验结果表明：这种基于相空间邻域的线性自适应预 测滤波器能够有效预测一些混沌序列，能够检测到一些混沌载波中的信号，具有去混沌噪声 的能力. 关键词： 混沌时间序列 重构矢量 自适应预测     

Long-term trends of tropospheric carbon monoxide and hydrogen cyanide from analysis of high resolution infrared solar spectra

Long-term trend and seasonal variation of the mean free tropospheric volume mixing ratios of carbon monoxide (CO) and hydrogen cyanide (HCN) have been derived from analysis of a time series of solar absorption spectra recorded from the US National Solar Observatory on Kitt Peak (31.9°N, 111.6°W, 2.09 km altitude) spanning almost three decades. The results of a fit to the CO 258 daily averages from May 1977 to April 2005 as a function of time with a model that assumes a sinusoidal seasonal cycle and a linear long-term trend with time yield a mean volume mixing ratio of 102±3) parts per billion (10^-9 per unit volume) below 10 km altitude, 1 sigma. The CO measurements show a seasonal cycle with a maximum in March and a minimum in September with an amplitude of (22.3±1.5)% relative to the mean. The best-fit corresponds to a long-term CO trend of , 1 sigma, relative to the mean. To quantify the possible impact of periods of intense fires, the CO measurements have been compared with the measurements of HCN, a well-documented emission product of biomass burning with a lifetime of ∼5 months. The best fit to the full HCN time series of 208 daily averages from May 1978 to April 2005 results in a mean volume mixing ratio of below 10 km altitude with a similar seasonal cycle, though with a lower relative amplitude than for CO. Although same-day enhancements up to a factor of 1.87 for HCN and 1.24 for CO were measured relative to values predicted by a fit to the time series that accounts for the seasonal cycles and trends of both molecules, excluding time periods of elevated fire emissions has no significant impact on the best-fit long-term free tropospheric CO and HCN trends. Our result of no long-term CO trend since the late 1970s suggests that the global average long-term decline reported from 1990 through 1995 measurements has not continued in the free troposphere. Similarly, a fit to the full time series of 208 HCN free tropospheric daily averages with the same model yields an average 2.09-10 km mixing ratio of 0.219 ppbv and a long-term trend of , 1 sigma, relative to the mean since 1978, also indicating no significant long-term trend above the lower mid-latitude continental US Kitt Peak station. The results for both molecules suggest the site was not significantly impacted by summer boreal fires during the time span of the measurements that in some years cause widespread pollution above northern higher latitude sites.     

Irregular vocal-fold vibration--high-speed observation and modeling

Direct observations of nonstationary asymmetric vocal-fold oscillations are reported. Complex time series of the left and the right vocal-fold vibrations are extracted from digital high-speed image sequences separately. The dynamics of the corresponding high-speed glottograms reveals transitions between low-dimensional attractors such as subharmonic and quasiperiodic oscillations. The spectral components of either oscillation are given by positive linear combinations of two fundamental frequencies. Their ratio is determined from the high-speed sequences and is used as a parameter of laryngeal asymmetry in model calculations. The parameters of a simplified asymmetric two-mass model of the larynx are preset by using experimental data. Its bifurcation structure is explored in order to fit simulations to the observed time series. Appropriate parameter settings allow the reproduction of time series and differentiated amplitude contours with quantitative agreement. In particular, several phase-locked episodes ranging from 4:5 to 2:3 rhythms are generated realistically with the model.     

The Fermi–Pasta–Ulam Problem and Its Underlying Integrable Dynamics: An Approach Through Lyapunov Exponents

FPU models, in dimension one, are perturbations either of the linear model or of the Toda model; perturbations of the linear model include the usual \(\beta \)-model, perturbations of Toda include the usual \(\alpha +\beta \) model. In this paper we explore and compare two families, or hierarchies, of FPU models, closer and closer to either the linear or the Toda model, by computing numerically, for each model, the maximal Lyapunov exponent \(\chi \). More precisely, we consider statistically typical trajectories and study the asymptotics of \(\chi \) for large N (the number of particles) and small \(\varepsilon \) (the specific energy E / N), and find, for all models, asymptotic power laws \(\chi \simeq C\varepsilon ^a\), C and a depending on the model. The asymptotics turns out to be, in general, rather slow, and producing accurate results requires a great computational effort. We also revisit and extend the analytic computation of \(\chi \) introduced by Casetti, Livi and Pettini, originally formulated for the \(\beta \)-model. With great evidence the theory extends successfully to all models of the linear hierarchy, but not to models close to Toda.     

Blind normalization of speech from different channels

We show how to construct a channel-independent representation of speech that has propagated through a noisy reverberant channel. This is done by blindly rescaling the cepstral time series by a nonlinear function, with the form of this scale function being determined by previously encountered cepstra from that channel. The rescaled form of the time series is an invariant property of it in the following sense: It is unaffected if the time series is transformed by any time-independent invertible distortion. Because a linear channel with stationary noise and impulse response transforms cepstra in this way, the new technique can be used to remove the channel dependence of a cepstral time series. In experiments, the method achieved greater channel-independence than cepstral mean normalization, and it was comparable to the combination of cepstral mean normalization and spectral subtraction, despite the fact that no measurements of channel noise or reverberations were required (unlike spectral subtraction).     

Bayer格式图像的实时彩色复原

为了减少算法的计算量,保证系统的实时性,本文针对Bayer格式图像提出了一种有效的彩色复原插值算法。插值过程中利用了人眼的视觉特性,能够更精确地得到图像的亮度信息和边缘信息。利用彩色图像的边缘特性更精确地复原了边缘处的R、G、B值。算法最终解为一系列5×5大小稀疏的线性滤波器,其复杂度低,实现简单,能在计算机各种嵌入式处理器中完成实时处理。实验证明,本算法的峰值信噪比(PSNR)比通常采用的双线性算法高4~6 db,且有效地减少了插值算法中出现的锯齿现象,使图像彩色的复原性和实时性比双线性算法更优越,具有一定的应用价值。     

基于局域相关向量机回归模型的小尺度网络流量的非线性预测

基于非线性时间序列局域预测法与相关向量机回归模型, 本文提出了局域相关向量机预测方法, 并应用于预测实际的小尺度网路流量序列. 应用基于信息准则的局域预测法邻近点的选取方法来选取局域相关向量机回归模型的邻近点个数. 对比分析了局域相关向量机预测法、前馈神经网络模型与局域线性预测法对网络流量序列的预测性能, 其中前馈神经网络模型的参数采用粒子群优化算法来优化. 实验结果表明: 邻近点优化后的局域相关向量机回归模型能够有效地预测小尺度网络流量序列, 归一化均方误差很小; 局域相关向量机回归模型生成的时间序列具有与原网络流量时间序列相一致的概率分布; 局域相关向量机回归模型的预测精度好于前馈神经网络模型的与局域线性预测法的. 关键词： 小尺度网络流量 非线性时间序列预测方法 局域预测法 相关向量机回归模型     

Correlations of record events as a test for heavy-tailed distributions

A record is an entry in a time series that is larger or smaller than all previous entries. If the time series consists of independent, identically distributed random variables with a superimposed linear trend, record events are positively (negatively) correlated when the tail of the distribution is heavier (lighter) than exponential. Here we use these correlations to detect heavy-tailed behavior in small sets of independent random variables. The method consists of converting random subsets of the data into time series with a tunable linear drift and computing the resulting record correlations.