Uniform Manifold Approximation and Projection Analysis of Soccer Players

In professional soccer, the choices made in forming a team lineup are crucial for achieving good results. Players are characterized by different skills and their relevance depends on the position that they occupy on the pitch. Experts can recognize similarities between players and their styles, but the procedures adopted are often subjective and prone to misclassification. The automatic recognition of players’ styles based on their diversity of skills can help coaches and technical directors to prepare a team for a competition, to substitute injured players during a season, or to hire players to fill gaps created by teammates that leave. The paper adopts dimensionality reduction, clustering and computer visualization tools to compare soccer players based on a set of attributes. The players are characterized by numerical vectors embedding their particular skills and these objects are then compared by means of suitable distances. The intermediate data is processed to generate meaningful representations of the original dataset according to the (dis)similarities between the objects. The results show that the adoption of dimensionality reduction, clustering and visualization tools for processing complex datasets is a key modeling option with current computational resources.     

基于正交投影散度的高光谱遥感波段选择算法

由于高光谱数据的海量高维特征,对其进行降维处理成为高光谱遥感研究的一个重要问题.波段选择算法由于能够有效地保留原始数据的信息,在高光谱数据降维及后续的遥感识别与分类等方面具有明显的优越性.文章提出了一种基于正交投影散度(OPD)的波段选择方法,该方法继承了正交子空间投影(OSP)算法的特点,通过把原始数据投影到特征空间...     

Collective signaling behavior in a networked-oscillator model

We propose and study the collective behavior of a model of networked signaling objects that incorporates several ingredients of real-life systems. These ingredients include spatial inhomogeneity with grouping of signaling objects, signal attenuation with distance, and delayed and impulsive coupling between non-identical signaling objects. Depending on the coupling strength and/or time-delay effect, the model exhibits completely, partially, and locally collective signaling behavior. In particular, a correlated signaling (CS) behavior is observed in which there exist time durations when nearly a constant fraction of oscillators in the system are in the signaling state. These time durations are much longer than the duration of a spike when a single oscillator signals, and they are separated by regular intervals in which nearly all oscillators are silent. Such CS behavior is similar to that observed in biological systems such as fireflies, cicadas, crickets, and frogs. The robustness of the CS behavior against noise is also studied. It is found that properly adjusting the coupling strength and noise level could enhance the correlated behavior.     

Visualization of 3D variable in time object based on data gathered by active measurement system

The recent needs of analysis and visualization of variable in time real 3D objects in many applications require development of new approach towards combining rapid 3D shape acquisition and the methodology of data processing in order to perform visualization and analysis of real 3D dynamic objects. In this paper, the general concept of visualization system of data gathered by means of optical 4D (x,y,z,t) shape measurement system is presented. The concept of a virtual camera, as the mean for interactive object visualization is introduced. The experimental results for processing of simulated and real variable in time 3D object are presented and discussed. The directions of future works focused on full implementation of the concept are introduced.     

激光诱导击穿光谱的自组织特征映射结合相关判别对天然地质样品分类方法研究

激光诱导击穿光谱技术具有微损、原位、快速分析的特点,在样品分类识别、成分分析等领域有广阔的应用前景。为探索该技术在天然地质样品识别应用的可行性,提出了一种自组织特征映射神经网络结合相关判别对天然地质样品LIBS光谱分类识别的方法。为减小全谱中背景噪声等不相关数据干扰、降低计算量,在元素谱线归属的基础上进行了特征谱线提取,实现了高维光谱数据的降维。以特征谱数据为输入建立网络训练模型,得到具有输入样本特征的权向量,通过权向量与待测样本进行相关分析可以实现样品分类。对16种天然地质样品的分类算法实验证明,在全谱、主成分降维和特征谱段三种数据处理方法中,特征谱的降维和提取LIBS数据主特征效果最优。改进的SOM网络结合相关判别算法比支持向量机方法和直接应用SOM网络方法的分类准确度更高,初步证实了该方法的有效性。     

Statistical properties, dynamic conditional correlation and scaling analysis: Evidence from Dow Jones and Nasdaq high-frequency data

This paper investigates statistical properties of high-frequency intraday stock returns across various frequencies. Both time series and panel data are utilized to explore the properties of probability distribution, dynamic conditional correlations, and scaling analysis in Dow Jones Industrial Average (DJIA) and Nasdaq intraday returns across 10-min, 30-min, 60-min, 120-min, and 390-min frequencies. The evidence shows that both returns and volatility (standard deviation) increase with the increasing scaling from 10-min to 390-min intervals. By fitting an AR(1)-GARCH(1,1) model to intraday data, we find that AR(1) coefficients are negative for DJIA returns and positive for Nasdaq, exhibiting a positive and negative feedback strategy in DJIA and Nasdaq, respectively. The evidence also shows that these coefficients are statistically significant for either including or excluding opening returns for the 10-min and 30-min frequencies. By examining the dynamic conditional correlation between the DJIA and the Nasdaq across different frequencies, a positive correlation ranging from 0.6 to 0.8 was found. In addition, the variance of the dynamic correlation coefficients is decreasing and appears to be stable for the 2001-2003 period. Finally, both returns on DJIA and Nasdaq satisfy the stable Lévy distributions, implying that both markets can converge to equilibrium by self-governing mechanism after shocks. Results of this work provide relevant implications for investors and policy makers.     

基于双向长短期记忆网络的太赫兹光谱识别

特征提取是太赫兹光谱识别的关键处理步骤,通常利用降维方法作为特征提取手段。然而,当一些化合物的太赫兹光谱曲线整体差异度较小时,降维方法往往会缺失样本差异的重要特征信息,从而导致分类错误。如果不采用降维方法提取特征,传统机器学习分类算法对维数较高的原始太赫兹光谱数据又不能很好的分类。针对此问题,提出了一种基于双向长短期记忆网络（BLSTM-RNN）自动提取太赫兹光谱特征的识别方法。BLSTM-RNN作为一种特殊的循环神经网络,利用其LSTM单元可以有效解决原始太赫兹光谱数据维数较高使得模型难以训练问题。再结合模型的双向频谱信息利用架构模式,可以增强模型对复杂光谱数据自动提取有效特征信息的能力。采用三类、15种化合物太赫兹透射光谱作为测试对象,首先利用S-G滤波和三次样条插值对Anthraquinone,Benomyl和Carbazole等十五种化合物在0.9～6 THz内的太赫兹透射光谱数据进行归一化处理,然后通过构建一个具有双向长短期记忆的循环神经网络对太赫兹光谱的全频谱信息进行自动特征提取并利用Softmax分类器进行分类。通过试验优化网络结构和各项参数,最终获得了针对复杂太赫兹透射光谱数据的预测模型,并与传统机器学习算法SVM,KNN及神经网络算法MLP,CNN进行对比实验。结果表明,dataset-1和dataset-2分别作为差异度较大和无明显峰值特征的五种化合物太赫兹透射光谱数据集,其平均识别率分别为100%和98.51%,与其他方法相比识别率有所提高;最重要的是,dataset-3作为5种化合物谱线极为相似的太赫兹透射光谱数据集,其平均识别率为96.56%,与其他方法相比识别率提高显著;dataset-4作为dataset-1,dataset-2和dataset-3的透射光谱数据集集合,其平均识别率为98.87%。从而验证了BLSTM-RNN模型能自动提取有效的太赫兹光谱特征,同时又能保证复杂太赫兹光谱的预测精度。在选择模型训练优化算法方面,使用Adam优化算法要好于RMSProp,SGD和AdaGrad,其模型的目标函数损失值收敛速度最快。同时随着模型训练迭代次数增加,相似太赫兹透射光谱数据集的预测准确率也不断提升。可为复杂太赫兹光谱数据库的光谱识别检索提供一种新的识别方法。     

Towards predictive combustion kinetic models: Progress in model analysis and informative experiments

One of the key tasks of combustion chemistry research is to develop accurate and robust combustion kinetic models for practical fuels. An accurate and robust kinetic model yields predictions that are highly consistent with experimental measurements over a wide range of operating conditions, with prediction uncertainties that are acceptable. Reliable experimental data generated by various powerful diagnostic techniques continue to play an essential role in the development of such models. This review focuses on the contributions of synchrotron-based species measurements in combustion systems, on model validation, model structure development, and model parameter optimization. Special emphasis is placed on recently reported strategies for informative and reliable experimental data generation, including combustion kinetic model input parameter evaluation, computational cost reduction for model analysis, model-analysis-based experimental design, experimental data treatment and error reduction. Particularly, the active-subspace-based method (ASSM) can reduce the dimensionality of combustion kinetic models and the aritificial-neural-network-based surrogates (ANN-HDMR and ANN-MCMC) can reduce the computational cost significantly. Global-sensitivity-based experimental design methods including sensitivity entropy and surrogate model similarity (SMS) can guide kinetics-information-enriched experimental data generation. Model-analysis-based calibration for experimental errors and feature extraction of experimental targets can improve the experimental data quality. A computational framework (OptEx) enabling the integration of experimental data with mechanism development, experimental design and model optimization, provides a new means to develop reliable kinetic models more efficiently and effectively.     

Advances in diversity profiling and combinatorial series design

Rapid advances in synthetic and screening technology have recently enabled the simultaneous synthesis and biological evaluation of large chemical libraries containing hundreds to tens of thousands of compounds, using molecular diversity as a means to design and prioritize experiments. This paper reviews some of the most important computational work in the field of diversity profiling and combinatorial library design, with particular emphasis on methodology and applications. It is divided into four sections that address issues related to molecular representation, dimensionality reduction, compound selection, and visualization.     

Scientific Review: New Software for Neutron Scattering Data Visualization

New software for neutron scattering data visualization, reduction, and analysis is being developed at IPNS through collaboration between IPNS and the University of Wisconsin-Stout. Recent advances in this Integrated Spectral Analysis Workbench (ISAW) package include: 1) a data analysis package for Single Crystal Diffractometers, 2) a data reduction package for small angle scattering, 3) wizards-a new method of operation which allows the user to step through a series of input forms to do a complete analysis, and 4) a newly designed set of viewers which work with basic data arrays and thus do not require the more complex data objects used for the first generation of viewers.     

Intracranial vessel wall imaging framework – Data acquisition,processing, and visualization

ObjectiveAssessment of vessel walls is an integral part in diagnosis and disease monitoring of vascular diseases such as vasculitis. Vessel wall imaging (VWI), in particular of intracranial arteries, is the domain of Magnetic Resonance Imaging (MRI) – but still remains a challenge. The tortuous anatomy of intracranial arteries and the need for high resolution within clinically acceptable scan times require special technical conditions regarding the hardware and software environments.Materials and methodsIn this work a dedicated framework for intracranial VWI is presented offering an optimized, black-blood 3D T1-weighted post-contrast Compressed Sensing (CS)-accelerated MRI sequence prototype combined with dedicated 3D-GUI supported post-processing tool for the CPR visualization of tortuous arbitrary vessel structures.ResultsUsing CS accelerated MRI sequence, the scanning time for high-resolution 3D black-blood CS-space data could be reduced to under 10 min. These data are adequate for a further processing to extract straightened visualizations (curved planar reformats – CPR). First patient data sets could be acquired in clinical environment.ConclusionA highly versatile framework for VWI visualization was demonstrated utilizing a post-processing tool to extract CPR reformats from high-resolution 3D black-blood CS-SPACE data, enabling simplified and optimized assessment of intracranial arteries in intracranial vascular disorders, especially in suspected intracranial vasculitis, by stretching their tortuous course. The processing time from about 15–20 min per patient (data acquisition and further processing) allows the integration into clinical routine.     

复杂场景下基于压缩感知的目标电磁散射与成像

研究了处于复杂场景下目标的逆合成孔径雷达(ISAR)成像问题。首先,建立了目标与复杂环境的电磁散射模型,采用计算电磁学的方法仿真得到了目标的雷达回波数据,进而充分考虑了背景噪声对雷达成像质量的影响。研究发现,目标所处的复杂背景会降低ISAR对目标的成像质量。其次,为减小仿真雷达回波数据所需的计算量,提出采用基于压缩感知(CS)的方法来对该场景进行成像,从而极大降低电磁仿真的计算点数。通过实验发现,在CS成像中,采用数据点使用率为0.4时所得到的成像质量可达到采用转台成像质量的效果。因此,采用基于CS的成像方法,可极大降低目标与场景的电磁散射计算复杂度,使得处于真实复杂场景下的目标电磁仿真和ISAR成像研究切实可行。     

Multifractal analysis of stock exchange crashes

We analyze the complexity of rare events of the DJIA Index. We reveal that the returns of the time series exhibit strong multifractal properties meaning that temporal correlations play a substantial role. The effect of major stock market crashes can be best illustrated by the comparison of the multifractal spectra of the time series before and after the crash. Aftershock periods compared to foreshock periods exhibit richer and more complex dynamics. Compared to an average crash, calculated by taking into account the larger 5 crashes of the DJIA Index, the 1929 event exhibits significantly more increase in multifractality than the 1987 crisis.     

三维荧光光谱结合稀疏主成分分析和支持向量机的油类识别方法研究

石油污染的出现,导致生态环境遭到破坏。因此,油类识别方法的研究对于环境的保护具有重要意义。采用荧光光谱法获得石油光谱数据,并对其进行预处理,再通过降维方法来提取特征信息,最后利用模式识别算法进行分类,从而可以实现对油类的定性分析,因此研究一种更高效的数据降维方法以及识别分类算法极其重要。基于三维荧光光谱技术,利用稀疏主成分分析（SPCA）对FS920光谱仪测得的荧光光谱数据进行特征提取,再利用支持向量机（SVM）算法对提取的特征数据进行分类识别,获得了一种更加高效的油类识别方法。首先,利用海水和十二烷基硫酸钠（SDS）配制成浓度为0.1 mol·L-1的胶束溶液,将其作为溶剂配制柴油、航空煤油、汽油以及润滑油各20种不同浓度的溶液;然后,利用FS920光谱仪测得样本溶液的三维荧光光谱数据,对得到的光谱数据进行预处理;最后,对预处理后的数据分别利用SPCA和主成分分析（PCA）进行特征提取,再利用SVM和K最近邻（KNN）两种模式识别算法对特征向量进行分类,最终得到四种模型PCA-KNN,SPCA-KNN,PCA-SVM以及SPCA-SVM的分类结果。研究结果表明,由四种模型得到的分类准确率分别为85%,90%,90%和95%,其中,在同种分类算法中,利用SPCA进行特征提取得到的分类准确率均比PCA的准确率高5%,因此可知,SPCA的稀疏性具有突出主要成分的作用,在提取光谱特征时能够减小非必要成分的影响,并且载荷矩阵的稀疏化可以去除变量之间的冗余信息,优化降维特征信息,为后续分类提供更有效的数据特征信息;在同种特征提取算法下,利用SVM算法进行分类得到的分类准确率均比KNN算法得到的准确率高5%,表明SVM算法在分类中更具有优势。因此,本文利用三维荧光光谱技术结合SPCA和SVM算法,实现了对石油的准确识别与分类,为今后对石油污染物的高效检测提供了新思路。     

Computational methods for electron tomography

Electron tomography (ET) has emerged as a powerful technique to address fundamental questions in molecular and cellular biology. It makes possible visualization of the molecular architecture of complex viruses, organelles and cells at a resolution of a few nanometres. In the last decade ET has allowed major breakthroughs that have provided exciting insights into a wide range of biological processes. In ET the biological sample is imaged with an electron microscope, and a series of images is taken from the sample at different views. Prior to imaging, the sample has to be specially prepared to withstand the conditions within the microscope. Subsequently, those images are processed and combined to yield the three-dimensional reconstruction or tomogram. Afterwards, a number of computational steps are necessary to facilitate the interpretation of the tomogram, such as noise reduction, segmentation and analysis of subvolumes. As the computational demands are huge in some of the stages, high performance computing (HPC) techniques are used to make the problem affordable in reasonable time. This article intends to comprehensively review the methods, technologies and tools involved in the different computational stages behind structural studies by ET, from image acquisition to interpretation of tomograms. The HPC techniques usually employed to cope with the computational demands are also briefly described.     

Dynamical criteria for the evolution of the stochastic dimensionality in flows with uncertainty

We estimate and study the evolution of the dominant dimensionality of dynamical systems with uncertainty governed by stochastic partial differential equations, within the context of dynamically orthogonal (DO) field equations. Transient nonlinear dynamics, irregular data and non-stationary statistics are typical in a large range of applications such as oceanic and atmospheric flow estimation. To efficiently quantify uncertainties in such systems, it is essential to vary the dimensionality of the stochastic subspace with time. An objective here is to provide criteria to do so, working directly with the original equations of the dynamical system under study and its DO representation. We first analyze the scaling of the computational cost of these DO equations with the stochastic dimensionality and show that unlike many other stochastic methods the DO equations do not suffer from the curse of dimensionality. Subsequently, we present the new adaptive criteria for the variation of the stochastic dimensionality based on instantaneous (i) stability arguments and (ii) Bayesian data updates. We then illustrate the capabilities of the derived criteria to resolve the transient dynamics of two 2D stochastic fluid flows, specifically a double-gyre wind-driven circulation and a lid-driven cavity flow in a basin. In these two applications, we focus on the growth of uncertainty due to internal instabilities in deterministic flows. We consider a range of flow conditions described by varied Reynolds numbers and we study and compare the evolution of the uncertainty estimates under these varied conditions.     

Improvement of 3D acquisition and visualization in MRI.

Three-dimensional (3D) visualization techniques are becoming an ever more important aid in the interpretation of tomographic data. Up to now, however, they have not received widespread use in MRI, because both acquisition and visualization techniques have been inadequate. In this paper we describe new 3D acquisition techniques which can acquire up to 128 slices with a resolution of 256 x 256 pixels in from 8 to 20 min. These techniques produce 3D data sets with excellent contrast and few motion artifacts, which are very well suited for 3D visualization techniques. For the visualization we investigate several rendering techniques, describe some improvements and compare their results. We found that there is no single method which renders all objects equally well. We show which shading method is best suited for different objects and why the other methods fail. Our studies suggest that in a 3D view with several objects each object should be rendered with a separate shading method. In so doing, 3D views can be generated which look like the real human anatomy.     

Comparing the structure of an emerging market with a mature one under global perturbation

In this paper we investigate the Tehran stock exchange (TSE) and Dow Jones Industrial Average (DJIA) in terms of perturbed correlation matrices. To perturb a stock market, there are two methods, namely local and global perturbation. In the local method, we replace a correlation coefficient of the cross-correlation matrix with one calculated from two Gaussian-distributed time series, whereas in the global method, we reconstruct the correlation matrix after replacing the original return series with Gaussian-distributed time series. The local perturbation is just a technical study. We analyze these markets through two statistical approaches, random matrix theory (RMT) and the correlation coefficient distribution. By using RMT, we find that the largest eigenvalue is an influence that is common to all stocks and this eigenvalue has a peak during financial shocks. We find there are a few correlated stocks that make the essential robustness of the stock market but we see that by replacing these return time series with Gaussian-distributed time series, the mean values of correlation coefficients, the largest eigenvalues of the stock markets and the fraction of eigenvalues that deviate from the RMT prediction fall sharply in both markets. By comparing these two markets, we can see that the DJIA is more sensitive to global perturbations. These findings are crucial for risk management and portfolio selection.     

Computer-generated kinoforms of real-existing full-color 3D objects using pure-phase look-up-table method

A computational method based on pure-phase look-up-table (LUT) is proposed for generating kinoforms of full-color real-existing three-dimensional (3D) objects. The principle of the pure-phase LUT method is described. 3D depth as well as color information of a full-color 3D object is obtained by the full-color 3D profile measurement approach based on binocular vision. The obtained full-color 3D data is decomposed into red (R), green (G), and blue (B) channels, and kinoforms of each channel are calculated from the depth map and color information of the corresponding channel using the pure-phase LUT method. In order to reduce the speckle noise of reconstructed full-color 3D image, sequential kinoforms of each color channel are generated by adding dynamic-pseudorandom phase factor into the object domain. Numerical reconstruction and optical reconstruction with a phase-only spatial light modulator (SLM) show that, with the proposed method, full-color holographic 3D display of real-existing full-color 3D objects is available.