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 ...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.展开更多
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...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.展开更多
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 p...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.展开更多
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 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.展开更多
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...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。展开更多
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 si...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.展开更多
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...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。展开更多
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...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.展开更多
文摘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.
基金Project supported in part by National Basic Research Program of China (973 Project) (Grant No 2006CB705506)Hi-Tech Research and Development Program of China (863 Project) (Grant No 2007AA11Z222)National Natural Science Foundation of China (Grant Nos 60721003 and 60774034)
文摘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.
文摘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.
文摘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.
文摘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。
基金Supported by the National Natural Science Foundation of China (No.10371053,10971249)
文摘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.
文摘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。
文摘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.
