Fluctuations and pseudo long range dependence in network flows: A non-stationary Poisson process model

In the study of complex networks （systems）, the scaling phenomenon of flow fluctuations refers to a certain powerlaw between the mean flux （activity） （Fi） of the i-th node and its variance σi as σi α （Fi）α Such scaling laws are found to be prevalent both in natural and man-made network systems, but the understanding of their origins still remains limited. This paper proposes a non-stationary Poisson process model to give an analytical explanation of the non-universal scaling phenomenon： the exponent α varies between 1/2 and 1 depending on the size of sampling time window and the relative strength of the external/internal driven forces of the systems. The crossover behaviour and the relation of fluctuation scaling with pseudo long range dependence are also accounted for by the model. Numerical experiments show that the proposed model can recover the multi-scaiing phenomenon.

The Impact of Long Range Dependence of VBR Video Traffic in Queueing System

he objective of the paper is to clarify the contradictory opinions about the effect of long range dependence (LRD) on the performance of queueing system. In order to investigate the impact of the value of LRD on the performance of queueing system, a particular source (Star War) is used, whose Pframe and Bframe show the same frame size distribution and the same autocorrelation structure (i.e. the similar SRD and LRD) and the different values. The “Shuffling Operation” changing the order of original time series in some way is used to remove either SRD or LRD. The conditions under which LRD or SRD makes the dominant impact on single server FIFO queue and the importance of the amount of LRD are found after a set of queueing experiments.

Network resource allocation attack detection with long range dependence

The approach of traffic abnormality detection of network resource allocation attack did not have reliable signatures to depict abnormality and identify them. However, it is crucial for us to detect attacks accurately. The technique that we adopted is inspired by long range dependence ideas. We use the number of packet arrivals of a flow in fixed-length time intervals as the signal and attempt to extend traffic invariant “self-similarity”. We validate the effectiveness of the approach with simulation and trace analysis.

Respective Roles of Short-and Long-Range Interactions in Protein Folding

A new method was presented to discuss the respective roles of short- and long-range interactions in protein folding. It's based on an off-lattice model, which is also being called as toy model. Simulated annealing algorithm was used to search its native conformation. When it is applied to analysis proteins 1agt and 1aho, we find that helical segment cannot fold into native conformation without the influence of long-range interactions. That's to say that long-range interactions are the main determinants in protein folding. Key words toy model - protein folding - simulated annealing algorithm - short and long range interactions CLC number O 242.28 - Q71 Foundation item: Supported by the National Natural Science Foundation of China((60301009)Biography: WANG Long-hui (1976-), female, Ph. D candidate, research direction: machine learning, bioinformatics.

Short-range action and long-range action of the electrostatic forces within atomic nuclei

We study the interaction forces in atomic nuclei based on our expressions for the electrostatic interaction between spheres of arbitrary radii and charges. We prove that at small distances the proton-neutron electrostatic attraction forces are short-range-acting and the proton-proton electrostatic repulsion forces are long-range-acting. We obtain that these forces are commensurate with the nuclear forces. The protonneutron electrostatic attraction forces and the proton-proton electrostatic repulsion forces at the same distance between nucleons differ in absolute value by about an order of magnitude. It follows that based on electromagnetic interactions the neutrons are the binding building blocks in nuclear structures.

Method of Dynamic VaR and CVaR Risk Measures Forecasting for Long Range Dependent Time Series on the Base of the Heteroscedastic Model

The paper proposes a new method of dynamic VaR and CVaR risk measures forecasting. The method is designed for obtaining the forecast estimates of risk measures for volatile time series with long range dependence. The method is based on the heteroskedastic time series model. The FIGARCH model is used for volatility modeling and forecasting. The model is reduced to the AR model of infinite order. The reduced system of Yule-Walker equations is solved to find the autoregression coefficients. The regression equation for the autocorrelation function based on the definition of a long-range dependence is used to get the autocorrelation estimates. An optimization procedure is proposed to specify the estimates of autocorrelation coefficients. The procedure for obtaining of the forecast values of dynamic risk measures VaR and CVaR is formalized as a multi-step algorithm. The algorithm includes the following steps: autoregression forecasting, innovation highlighting, obtaining of the assessments for static risk measures for residuals of the model, forming of the final forecast using the proposed formulas, quality analysis of the results. The proposed method is applied to the time series of the index of the Tokyo stock exchange. The quality analysis using various tests is conducted and confirmed the high quality of the obtained estimates.

STATISTICAL PROPERTIES OF LONG-RANGE CONTACTS IN GLOBULAR PROTEINS

The analysis of residue-residue contacts in protein structures can shed some light on our understanding of the folding and stability of proteins. In this paper, we study the statistical properties of long-range and short-range residue- residue contacts of 91 globular proteins using CSU software and analyze the importance of long-range contacts in globular protein structure. There are many short-range and long-range contacts in globular proteins, and it is found that the average number of long-range contacts per residue is 5.63 and the percentage of residue-residue contacts which are involved in long- range ones is 59.4%. In more detail, the distribution of long-range contacts in different residue intervals is investigated and it is found that the residues occurring in the interval range of 4-10 residues apart in the sequence contribute more long-range contacts to the stability of globular protein. The number of long-range contacts per residue, which is a measure of ability to form residue-residue contacts, is also calculated for 20 different amino acid residues. It is shown that hydrophobic residues (including Leu, Val, He, Met, Phe, Tyr, Cys and Trp) having a large number of long-range contacts easily form long-range contacts, while the hydrophilic amino acids (including Ala, Gly, Thr, His, Glu, Gln, Asp, Asn, Lys, Ser, Arg, and Pro) form long-range contacts with more difficulty. The relationship between the Fauchere-Pliska hydrophobicity scale (FPH) and the number of short-range and long-range contacts per residue for 20 amino acid residues is also studied. An approximately linear relationship between the Fauchere-Pliska hydrophobicity scale (FPH) and the number of long-range contacts per residue CL, is found and can be expressed as CL = a + b × FPH where a = 5.04 and b = 1.23. These results can help us to understand the role of residue-residue contacts in globular protein structure.

基于长记忆视角的Granger因果检验

基于ARFIMA和FIGARCH模型,以地产股指数和银行股指数为例,提供了一种分解长记忆时间序列短记忆特征的思路,研究了时间序列长记忆性对Granger因果检验的影响。案例研究表明,银行指数收益率波动对地产指数收益率波动有更强的因果引致关系,且银行指数收益率波动的长记忆特征在一定程度上影响到其短时相依关系对地产指数收益率波动的解释力。

Heavy Traffic Limit Theorems for a Queue with Poisson ON/OFF Long-range Dependent Sources and General Service Time Distribution

In Internet environment, traffic flow to a link is typically modeled by superposition of ON/OFF based sources. During each ON-period for a particular source, packets arrive according to a Poisson process and packet sizes （hence service times） can be generally distributed. In this paper, we establish heavy traffic limit theorems to provide suitable approximations for the system under first-in first-out （FIFO） and work-conserving service discipline, which state that, when the lengths of both ON- and OFF-periods are lightly tailed, the sequences of the scaled queue length and workload processes converge weakly to short-range dependent reflecting Gaussian processes, and when the lengths of ON- and/or OFF-periods are heavily tailed with infinite variance, the sequences converge weakly to either reflecting fractional Brownian motions （FBMs） or certain type of long- range dependent reflecting Gaussian processes depending on the choice of scaling as the number of superposed sources tends to infinity. Moreover, the sequences exhibit a state space collapse-like property when the number of sources is large enough, which is a kind of extension of the well-known Little＇s law for M/M/1 queueing system. Theory to justify the approximations is based on appropriate heavy traffic conditions which essentially mean that the service rate closely approaches the arrival rate when the number of input sources tends to infinity.

WAVELET-BASED ESTIMATORS OF MEAN REGRESSION FUNCTION WITH LONG MEMORY DATA

This paper provides an asymptotic expansion for the mean integrated squared error （MISE） of nonlinear wavelet-based mean regression function estimators with long memory data. This MISE expansion, when the underlying mean regression function is only piecewise smooth, is the same as analogous expansion for the kernel estimators.However, for the kernel estimators, this MISE expansion generally fails if the additional smoothness assumption is absent.

热和低温处理对淀粉长短程有序结构的影响及变化机理研究进展

淀粉中长短程有序结构的紧密排列、高度有序能使其表现出良好的热稳定性和抗酶解能力,因此淀粉处理过程中长短程有序结构的改变对其性质尤其是消化性能的影响成为研究热点之一。热和低温处理是淀粉基食品常见的处理方式,主要分别通过热效应和生成冰晶产生挤压破坏淀粉分子,断裂的淀粉结构随游离水的流动重排会影响淀粉长短程有序结构的含量和完整性,从而影响淀粉的糊化、流变、消化等性质。文中以湿热、韧化、干热、冻融、冻干这5种处理方式为主,综述了热和低温对淀粉长短程有序结构的影响,并从精细结构变化的角度阐述了影响机理,同时指出长短程有序结构的变化及其对淀粉理化及功能性质的影响,可为功能性淀粉基食品的开发提供思路和参考。

长相关随机模型FBM对光伏发电短期预测

针对光伏发电功率具有较强的波动性、间歇性输出,光伏功率预测精度较低,且难于给出具体预测时间长度等问题,提出了一种长相关随机模型分数阶布朗运动(fractional Brownian motion,FBM),用于光伏功率预测。首先,采用重标极差法计算长相关(long-range dependence,LRD)参数-Hurst指数,Hurst指数用于判断光伏功率数据是否满足长相关性,并通过最大李雅普诺夫指数(Lyapunov)计算出模型最大可预测时间尺度;其次,采用随机微分法建立FBM光伏功率预测模型,同时估计FBM预测模型参数值;最后,选取澳大利亚沙漠知识太阳能中心(Desert Knowledge Australia Solar Center,DKASC)、美国国家可再生能源实验室(National Renewable Energy Laboratory,NREL)以及北京国能日新科技有限公司的光伏功率数据集,从不同的地理环境、不同的气候特征、不同的规模大小电站进行验证。仿真结果表明,该模型较传统的Kalman、LSTM模型具有更高的预测精度,可为光伏并网的稳定和安全运行提供更好的理论支持,对电网调度部门具有较高的参考价值。

联合句法与位置信息的方面情感三元组抽取

为提高方面级情感三元组抽取任务的准确率,提出一种联合依存句法关系和位置偏移信息的抽取模型。在模型上下文编码中添加句法关系,结合图卷积网络捕获结构和结点属性信息,增强三元组要素之间的交互能力;在多任务学习部分加入相对位置偏移信息,充分挖掘方面-观点词对的关系,提高三元组要素抽取的精度。在4个基准英文数据集上的实验结果表明,该方法效果显著且优于其它基线模型。

基于LSTM车速预测和深度确定性策略梯度的增程式电动汽车能量管理

为提高增程式电动汽车的能量管理性能,首先利用长短时记忆(LSTM)神经网络进行车速预测,然后计算出预测时域内的需求功率,并将其与当前时刻的需求功率共同输入深度确定性策略梯度(DDPG)智能体,由智能体输出控制量,最后通过硬件在环仿真验证了控制策略的实时性。结果表明,采用所提出的LSTM-DDPG能量管理策略相对于DDPG能量管理策略、深度Q网络(DQN)能量管理策略、功率跟随控制策略在世界重型商用车辆瞬态循环(WTVC)工况下的等效燃油消耗量分别减少0.613 kg、0.350 kg、0.607 kg,与采用动态规划控制策略时的等效燃油消耗量仅相差0.128 kg。

对比约束下的非局部关联单图像去反光级联算法

图像中存在的反光不但显著降低了图像质量,而且严重影响了后续的计算机视觉任务。因此,提出了一种对比约束下的非局部关联单图像去反光级联算法NCRR,该算法通过LSTM传递跨级联信息的双支路方式,利用反光特征和背景特征相互补充信息并迭代细化预测精度,使2条支路的预测效果相互促进。为了便于多个级联步骤的训练,提出了一种正负对比的正则化损失,将背景图像和原图像的特征分别作为正、负样本,确保目标图像在表示空间中拉近背景图像,推远原图像,缩小预测范围,较好地缓解不适定性问题。此外,提出了一种高效、计算量少的非局部关联预测模块,它能获取十字交叉路径上所有像素的上下文信息。通过进一步级联操作,使每个像素捕获整幅图像长距离的依赖关系,能利用周围像素点信息来预测被强反光遮挡的背景信息。实验结果表明,本文算法能够有效去除玻璃的强反光,并且玻璃去反光的评估结果都超过了其他对比算法,具有较好的鲁棒性。

壳聚糖添加量对大米淀粉老化的影响

为探究壳聚糖添加量对大米淀粉老化性质的影响,采用差示扫描量热仪、红外光谱(FTIR)和X射线衍射仪对添加不同添加量壳聚糖的老化大米淀粉进行测试,以老化大米淀粉的回生焓值、R值以及相对结晶度等3项指标来评价壳聚糖添加量对大米淀粉老化的影响。结果表明:添加壳聚糖使老化大米淀粉回生焓值增大;FTIR分析结果证实了添加壳聚糖使老化大米淀粉的短程有序性降低;壳聚糖添加量为0.5g(占淀粉干重3.3%)时,淀粉短期回生阶段的短程有序性最低,添加量为0.4g(占淀粉干重2.7%)时,淀粉长期回生阶段的短程有序性最低;老化大米淀粉结构的长程有序性同壳聚糖添加量有关:壳聚糖添加量为0.1g(占淀粉干重0.7%)时,壳聚糖能降低淀粉的相对结晶度,添加量增加到0.5g(占淀粉干重3.3%)后,壳聚糖增大淀粉的相对结晶度,且老化时间越长,使淀粉相对结晶度增大所需的壳聚糖添加量越大。研究结果有助于拓宽壳聚糖的应用范围,也为淀粉老化的调控提供了新思路。

横截面独立的长相依面板数据的公共均值变点

研究横截面独立的长相依面板数据的公共均值变点的估计问题.文中分别在强变点信号,中等变点信号和弱变点信号这三种情形下研究了公共变点的估计量,建立了估计量的渐近性质,包括相合性,收敛速度和极限分布.理论结果表明在变点信号和长相依之间存在着一种权衡关系.具体来说,在强变点信号情形下,长相依不影响估计量的渐近性质;在中等变点信号情形下,长相依不影响估计量的收敛速度,但影响估计量的极限分布;在弱变点信号情形下,长相依既影响估计量的收敛速度,也影响估计量的极限分布.Monte Carlo模拟评估了估计量在有限样本情形下的表现,并支持文中的理论结果.

基于电压极差特征的储能电池组早期健康状态检测

为了更加高效地评估储能电池组的健康状态(SOH),提出一种基于电压极差特征的早期健康状态检测方法。首先基于大容量磷酸铁锂储能电池组开展循环老化试验,测量每次循环的电压极差信号,并从中提取关键时间点的电压特征;其次,基于皮尔逊(Pearson)相关系数及灰色关联度分析法(GRA)筛选与电池组老化高度相关的健康因子。最后,通过麻雀搜索算法(SSA)优化双向长短时记忆网络(BiLSTM)的超参数,搭建SSA-BiLSTM健康状态估计模型,实现储能电池组SOH评估;并结合常规机器学习算法验证了健康因子的有效性和估计模型的优越性。结果表明,所提取充放电静置30 min的电压极差特征能够有效反映电池组容量衰退趋势,多种模型验证下SOH估计误差均低于±0.8%。其中,本文所提出的SSA-BiLSTM模型均方根误差(RMSE)低至0.07%。因此该方法能够有效地对大容量储能电池组的SOH实现在线监测。

结合多尺度注意力机制和双向门控循环网络的视频摘要模型

任务中全局注意力在长距离视频序列上注意力值分布的方差较大,生成关键帧的重要性分数偏差较大,且时间序列节点边界值缺乏长程依赖导致的片段语义连贯性较差等问题,通过改进注意力模块,采用分段局部自注意力和全局自注意力机制相结合来获取局部和全局视频序列关键特征,降低注意力值的方差。同时通过并行地引入双向门控循环网络(bidirectional recurrent neural network,BiGRU),二者的输出分别输入到改进的分类回归模块后再将结果进行加性融合,最后利用非极大值抑制(non-maximum suppression,NMS)和核时序分割方法(kernel temporal segmentation,KTS)筛选片段并分割为高质量代表性镜头,通过背包组合优化算法生成最终摘要,从而提出一种结合多尺度注意力机制和双向门控循环网络的视频摘要模型(local and global attentions combine with the BiGRU,LG-RU)。该模型在TvSum和SumMe的标准和增强数据集上进行了对比试验,结果表明该模型取得了更高的F-score,证实了该视频摘要模型保持高准确率的同时可鲁棒地对视频完成摘要。

孪生Swin Transformer的红外与可见光图像融合算法

为解决基于卷积神经网络(convolutional neural networks,CNN)的红外与可见光图像融合算法融合过程中未考虑原图像远程依赖关系的问题,提出了一种孪生Swin Transformer的红外与可见光图像融合算法,构建了孪生网络模型。使用潜在低秩表示(latent low-rank representation,LatLRR)分解方法将原图像分解,再分别融合以提升图像融合的精度;使用Swin Transformer分别提取分解后图像的特征,获取图像的远程依赖关系;使用l1-norm正则化方法求解特征范数,利用Softmax分别获得分解后图像的权重图;利用线性加权方法重构融合图像,通过简单的线性加和获得最终融合图像。在评测基准VIFB上对所提算法进行评测,实验结果表明,所提算法与20种融合算法相比,定性性能优良;定量分析,在由13种评价指标组成的评价体系中,该算法能够取得3个最优值,超过大部分融合算法;在运行时间方面,该算法的运行时间成本低于大部分融合算法。所提融合算法相较于其他融合算法,在主观和客观方面均表现出更好的融合性能。