Enhanced Temporal Correlation for Universal Lesion Detection

Universal lesion detection(ULD)methods for computed tomography(CT)images play a vital role in the modern clinical medicine and intelligent automation.It is well known that single 2D CT slices lack spatial-temporal characteristics and contextual information compared to 3D CT blocks.However,3D CT blocks necessitate significantly higher hardware resources during the learning phase.Therefore,efficiently exploiting temporal correlation and spatial-temporal features of 2D CT slices is crucial for ULD tasks.In this paper,we propose a ULD network with the enhanced temporal correlation for this purpose,named TCE-Net.The designed TCE module is applied to enrich the discriminate feature representation of multiple sequential CT slices.Besides,we employ multi-scale feature maps to facilitate the localization and detection of lesions in various sizes.Extensive experiments are conducted on the DeepLesion benchmark demonstrate that thismethod achieves 66.84%and 78.18%for FS@0.5 and FS@1.0,respectively,outperforming compared state-of-the-art methods.

Employing temporal channel correlation in user selection for MIMO broadcast systems

Taking the time varying nature of wireless channels into account, two user selection schemes with lower complexity are developed for multiple-input multiple-output broadcast （MIMO BC）systems. According to the relationship between coherence time and Doppler frequency, an information frame is divided into several segments. At the beginning of each segment, the user selection is carded out with the greedy selection algorithm. In the simplified user selection algorithms employing the temporal correlation （SUSTC）, the selection results are applied for all the remaining slots in each segment. But in the improved simplified user selection algorithms employing the temporal correlation（ISUSTC）, at the remaining slots, users are kept with favorable channel conditions selected at the previous slot, and other users are updated from the candidate pool to communicate simultaneously. Simulations show that compared with the greedy user selection method, the proposed algorithms can reduce the selection complexity with a little sum capacity loss.

Quantitative assessment of the spatio-temporal correlations of seismic events induced by longwall coal mining

Rock failure process as a natural response to mining activities is associated with seismic events, which can pose a potential hazard to mine operators, equipment and infrastructures. Mining-induced seismicity has been found to be internally correlated in both time and space domains as a result of rock fracturing during progressive mining activities. Understanding the spatio-temporal(ST) correlation of mininginduced seismic events is an essential step to use seismic data for further analysis, such as rockburst prediction and caving assessment. However, there are no established methods to perform this critical task. Input parameters used for the prediction of seismic hazards, such as the time window of past data and effective prediction distance, are determined based on site-specific experience without statistical or physical reasons to support. Therefore, the accuracy of current seismic prediction methods is largely constrained, which can only be addressed by quantitively assessing the ST correlations of mininginduced seismicity. In this research, the ST correlation of seismic event energy collected from a study mine is quantitatively analysed using various statistical methods, including autocorrelation function(ACF), semivariogram and Moran’s I analysis. In addition, based on the integrated ST correlation assessment, seismic events are further classified into seven clusters, so as to assess the correlations within individual clusters. The correlation of seismic events is found to be quantitatively assessable, and their correlations may vary throughout the mineral extraction process.

Effects of internal waves on signal temporal correlation length in the South China Sea

We simulated the temporal correlation of sound transmission using a two-dimensional advective frozen-ocean model with temperature data from a temperature sensor array on a propagation path in the South China Sea （SCS） Experiment 2009, and investigated the relationships of temporal correlation length, source-receiver range, and maximal sound speed fluctuation mainly caused by the solitary internal waves. We found that the temporal correlation length is -h2-power dependent on source-receiver range and -0.9-power dependent on maximal sound speed fluctuation. The empirical relationship is deduced from one-day environmental measurements in a limited area, needing more works and verification in the future with more acoustic data. But the relationship is useful in many applications in the area of SCS Experiment 2009.

Spatiotemporal Prediction of Urban Traffics Based on Deep GNN

Traffic prediction already plays a significant role in applications like traffic planning and urban management,but it is still difficult to capture the highly non-linear and complicated spatiotemporal correlations of traffic data.As well as to fulfil both long-termand short-termprediction objectives,a better representation of the temporal dependency and global spatial correlation of traffic data is needed.In order to do this,the Spatiotemporal Graph Neural Network(S-GNN)is proposed in this research as amethod for traffic prediction.The S-GNN simultaneously accepts various traffic data as inputs and investigates the non-linear correlations between the variables.In terms of modelling,the road network is initially represented as a spatiotemporal directed graph,with the features of the samples at the time step being captured by a convolution module.In order to assign varying attention weights to various adjacent area nodes of the target node,the adjacent areas information of nodes in the road network is then aggregated using a graph network.The data is output using a fully connected layer at the end.The findings show that S-GNN can improve short-and long-term traffic prediction accuracy to a greater extent;in comparison to the control model,the RMSE of S-GNN is reduced by about 0.571 to 9.288 and the MAE(Mean Absolute Error)by about 0.314 to 7.678.The experimental results on two real datasets,Pe MSD7(M)and PEMS-BAY,also support this claim.

Efficient User Identity Linkage Based on Aligned Multimodal Features and Temporal Correlation

User identity linkage(UIL)refers to identifying user accounts belonging to the same identity across different social media platforms.Most of the current research is based on text analysis,which fails to fully explore the rich image resources generated by users,and the existing attempts touch on the multimodal domain,but still face the challenge of semantic differences between text and images.Given this,we investigate the UIL task across different social media platforms based on multimodal user-generated contents(UGCs).We innovatively introduce the efficient user identity linkage via aligned multi-modal features and temporal correlation(EUIL)approach.The method first generates captions for user-posted images with the BLIP model,alleviating the problem of missing textual information.Subsequently,we extract aligned text and image features with the CLIP model,which closely aligns the two modalities and significantly reduces the semantic gap.Accordingly,we construct a set of adapter modules to integrate the multimodal features.Furthermore,we design a temporal weight assignment mechanism to incorporate the temporal dimension of user behavior.We evaluate the proposed scheme on the real-world social dataset TWIN,and the results show that our method reaches 86.39%accuracy,which demonstrates the excellence in handling multimodal data,and provides strong algorithmic support for UIL.

Temporal patterns of dengue epidemics:The case of recent outbreaks in Kaohsiung

Objective:To investigate whether major dengue outbreaks in the last two decades in Kaohsiung follow a precise temporal pattern.Methods:Government daily lab-confirmed dengue case data from three major dengue outbreaks occurring during the last two decades in Kaohsiung in2002,2014 and 2015,is utilized to compute the corresponding weekly cumulative percentage of total case numbers.We divide each of the three time series data into two periods to examine the corresponding weekly cumulative percentages of case numbers for each period.Pearson’s correlation coefficient was calculated to compare quantitatively the similarity between the temporal patterns of these three years.Results:Three cutoff points produce the most interesting comparisons and the most different outcomes.Pearson’s correlation coefficient indicates quantitative discrepancies in the similarity between temporal patterns of the three years when using different cutoff points.Conclusions:Temporal patterns in 2002 and 2014 are comparatively more similar in early stage.The 2015 outbreak started late in the year,but ended more like the outbreak in 2014,both with record-breaking number of cases.The retrospective analysis shows that the temporal dynamics of dengue outbreaks in Kaohsiung can strongly vary from one year to another,making it difficult to identify any common predictor.

Exploiting Correlations of Energy and Information:A New Paradigm of Energy Harvesting Communications

For deployment flexibility and device lifetime prolonging,energy harvesting communications have drawn much attention recently,which however,encounter energy domain randomness in addition to the channel state randomness and traffic load randomness.The three-dimensional randomness makes the resource allocation problem extremely difficult.To resolve this,we exploit the inherent correlations of energy arrival and information.The correlations include self correlations of energy profiles and mutual correlations between energy and information in both time and spatial domains.The correlations are explicitly explained followed by a state-of-art survey.Candidate mechanisms exploiting the correlations for the ease of resource allocation are introduced along with some recent progress.Finally,a case study is presented to illustrate the performance of the proposed algorithm.

The spatial-temporal evolution of coherent structures in log law region of turbulent boundary layer

The spatial-temporal evolution of coherent structures （CS） is significant for turbulence control and drag re- duction. Among the CS, low and high speed streak structures show typical burst phenomena. The analysis was based on a time series of three-dimensional and three-component （3D-3C） velocity fields of the flat plate turbulent boundary layer （TBL） measured by a Tomographic and Time-resolved PIV （Tomo TRPIV） system. Using multi-resolution wavelet transform and conditional sampling method, we extracted the intrinsic topologies and found that the streak structures appear in bar-like patterns. Furthermore, we seized locations and velocity information of transient CS, and then calculated the propagation velocity of CS based on spatial-temporal cross-correlation scanning. This laid a foundation for further studies on relevant dynamics properties.

Signal Analysis of the Climate: Correlation, Delay and Feedback

One of the ingredients of anthropogenic global warming is the existence of a large correlation between carbon dioxide concentrations in the atmosphere and the temperature. In this work we analyze the original time-series data that led to the new wave of climate research and test the two hypotheses that might explain this correlation, namely the (more commonly accepted and well-known) greenhouse effect (GHE) and the less-known Henry’s Law (HL). This is done by using the correlation and the temporal features of the data. Our conclusion is that of the two hypotheses the greenhouse effect is less likely, whereas the Henry’s Law hypothesis can easily explain all effects. First the proportionality constant in the correlation is correct for HL and is about two orders of magnitude wrong for GHE. Moreover, GHE cannot readily explain the concurring methane signals observed. On the temporal scale, we see that GHE has difficulty in the apparent negative time lag between cause and effect, whereas in HL this is of correct sign and magnitude, since it is outgasing of gases from oceans. Introducing feedback into the GHE model can overcome some of these problems, but it introduces highly instable and chaotic behavior in the system, something that is not observed. The HL model does not need feedback.

Two new approaches for image registration based onspatial-temporal relationship

How to improve the probability of registration and precision of localization is a hard problem, which is desiderated to solve. The two basic approaches (normalized cross-correlation and phase correlation) for image registration are analysed, two improved approaches based on spatial-temporal relationship are presented. This method adds the correlation matrix according to the displacements in x- cirection and y- directions, and the registration pose is searched in the added matrix. The method overcomes the shortcoming that the probability of registration decreasing with area increasing owing to geometric distortion, improves the probability and the robustness of registration.

Detecting remotely triggered temporal changes around the Parkfield section of the San Andreas fault

Detecting temporal changes in fault zone properties at seismogenic depth have been a long-sought goal in the seismological community for many decades. Recent studies based on waveform analysis of repeating earthquakes have found clear temporal changes in the shallow crust and around active fault zones associated with the occurrences of large nearby and teleseismic earthquakes. However, repeating earthquakes only occur in certain locations and their occurrence times cannot be controlled, which may result in inadequate sampling of the interested regions or time periods. Recent developments in passive imaging via auto- and cross-correlation of ambient seismic wavefields （e.g., seismic noise, earthquake coda waves） provide an ideal source for continuous monitoring of temporal changes around active fault zones. Here we conduct a systematic search of temporal changes along the Parkfield section of the San Andreas fault by cross-correlating relatively high-frequency （0.4-1.3 Hz） ambient noise signals recorded by 10 borehole stations in the High Resolution Seismic Network. After using stretch/compressed method to measure the delay time and the decorrelation-index between the daily noise cross-correlation functions （NCCFs）, we find clear temporal changes in the median seismic velocity and decorrelation-index associated with the 2004 M6.0 Parkfield earthquake. We also apply the same procedure to the seismic data around five regional/teleseismic events that have triggered non-volcanic tremor in the same region, but failed to find any clear temporal changes in the daily NCCFs. The fact that our current technique can detect temporal changes from the nearby but not regional and teleseismic events, suggests that temporal changes associated with distance sources are very subtle or localized so that they could not be detected within the resolution of the current technique （-0.2%）.

Temporal sequential correlation analysis on compoundaction potentials of the auditory nerve versus the stimulating AM-sound envelope

Compound action potentials of the auditory nerve in response to amplitude modulating tones were recorded in guinea pigs with electrode implanted to the exit of the internal auditory meatus and temporal sequential correlation between the responses and the modulators was studied in a paradigm of systematically changing acoustic parameters. Three kinds of modulators were used. continuous or burst sinusoids of fixed frequency (in the range of 40 Hz-5 kHz), short bursts of sinusoids with changing frequency and short segments of speech signal. Ranges of parametric variation were 500 Hz-20 kHz for carrier frequency, 5%-95% for modulation depth and 20 dB-90 dB SPL for intensity. For continuous or burst sinusoidal modulators of fixed frequencies, the correlation coefficient (r) remained quite high in most parametric conditions, ranging from 0.80 to 0.95. It became smaller mainly in instances of decreased response amplitude on account of unfavourable parameters. For burst modulators of changing frequency, r varied around 0.66-0.86. When segments of speech signal served as the modulators, significant correlation (r around 0.50 ) also existed, indicating the validity of the timing mode of information encoding for speech sound at the cochlear nerve level. Some theoretical and technical points in studying the timing mechanism of audition is discussed.

AMOUNT OF INFORMATION FOR SPATIAL-TEMPORAL CORRELATION FILTER PATTERN RECOGNITION SYSTEM

In order to relate the design and analysis of an optical pattern recognition system with the structural parameters, only by introducing the prolate spheroidal wave function (PSWF) can the amount of information be computed. Combining the imaging wave function set {ψi(x)} and distorted wave function set {b_i(p)} and two integral equations they satisfy derives the expression of the amount of information. The design method of matched filter connected with its amount of information is studied, and their amounts of information belonging to different pattern recognition systems are illustrated. It can be seen that the difference of the amounts of information for various systems is obvious.

基于时空关联特征与B-LSTM模型的分布式光伏功率区间预测

提出一种基于时空关联特征与贝叶斯-长短期记忆神经网络(bayesian long short-term memory,B-LSTM)模型的分布式光伏功率区间预测方法。以长短期记忆神经网络(long short-term memory,LSTM)为基础构建近似贝叶斯神经网络,建立考虑时空关联特征的B-LSTM模型,利用其强大的记忆能力和特征提取不同特征尺度的模态分量,并进行分布式光伏功率区间预测。以某地区实际分布式光伏数据集进行算例分析,验证了所提方法的优越性。

1960-2020年河南省极端降水时空演变特征

全球变暖加快了水文循环速度,导致极端降水事件频发,增强了城市排水负荷和洪涝风险,并影响了区域气候的时空特征。在河南省均匀选取26个雨量站对1960—2020年61年间的降水资料进行研究,采用趋势分析、Kriging插值、M-K突变分析、Morlet小波变换和Spearman相关系数等方法对河南省9个极端降水指数的时空演变特征进行分析。结果显示:河南省年降水量呈上升趋势,极端降水事件呈下降趋势,河南省整体气候逐渐变得湿润。河南省极端降水指数具有显著空间差异,东南部极端降水风险显著大于西北部。除降水强度、最大连续降水日数和1 d最大降水量外,其余指数与年降水量、汛期和6—9月的各月降水量具有良好的相关性,这对于评估年降水量及其年内分布特征、极端降水事件频率有较好的参考作用。结果表明:降水从短历时高强度向长历时低强度演变,要应对此类长历时极端降水的风险,防涝思路应从提高短时间大量排水能力转变到提高长时间低强度吸收降水能力。

冷能梯级利用的港口多能微网双层不确定性经济调度

为有效挖掘港口液化天然气(LNG)冷能利用的低碳灵活性潜力,充分发挥多时间尺度协同优化效应,提出一种考虑LNG冷能梯级利用的港口多能微网(MEMG)鲁棒-随机双层不确定性经济调度模型。首先,考虑LNG深冷-中冷-浅冷等各个温区的低碳灵活性潜力,建立低温碳捕集-冷能发电-直接冷却的冷能梯级利用模型,并以此为基础形成捕集-存储-利用协同的碳处理流程。其次,根据等概率逆变换生成考虑预测误差时序相关性的风电场景,并基于Wasserstein距离的0-1规划模型进行场景削减。再次,针对风电预测误差随时间尺度增加而增大的特性,构建多时间尺度优化的鲁棒-随机双层不确定性经济调度模型,上层通过分布鲁棒优化保证日前预调度决策鲁棒性,下层通过随机优化保证日内滚动调度决策经济性。最后,仿真结果表明,所提考虑冷能梯级利用的鲁棒-随机双层调度模型在解决日前长时间尺度预测精度低与日内短时间尺度易陷入局部最优矛盾的同时,可赋予港口MEMG更多经济性、低碳性及供电灵活性。

中国沿海地区植被NDVI时空变化及驱动力分析

研究植被变化对区域生态修复具有重要意义。以中国沿海地区为例,基于归一化植被指数(NDVI)和降水、夜间灯光等自然和人为因子数据,运用Theil-Sen Median趋势分析+Mann-Kendall检验、最优参数地理探测器(OPGD)、相关分析和Hurst指数,多时空尺度探讨了中国沿海地区植被NDVI时空变化规律及其驱动力。结果表明:1)2001-2020年研究区植被状况较好,NDVI多年均值为0.762,具体到各分区,东北沿海的NDVI均值最高,其次是华南沿海,华东沿海和华北沿海;全区NDVI逐年变化率为0.019/10 a(P<0.01),不同分区的上升趋势从大到小为华南沿海、东北沿海、华北沿海和华东沿海,区域内植被状况不断改善,退耕还林还草和沿海防护林等生态工程效益不断显现;2)夜间灯光指数在全区各个因子中的解释力最大(q值为0.354),人为因素对NDVI的解释力明显大于自然因素,其对植被恢复产生了积极影响,并且随时间推移逐渐增强;3)两因子结合后的解释力大于单因子,表现为双因子增强和非线性增强。在全区范围内,影响最大的一对交互作用为土壤类型∩夜间灯光,其他分区则为日照时数∩夜间灯光(东北沿海地区),土壤类型∩夜间灯光(华北沿海和华东沿海地区),人口密度∩夜间灯光(华南沿海地区),自然因素和人类活动因素作用后影响力有了显著提升,但人类活动因素仍占据主导地位;4)Hurst指数均值为0.463,未来一段时间内,研究区内植被变化有66.3%的地区表现出一定的反持续性。研究结果有利于为中国沿海地区生态保护和高质量发展提供科学支撑。

基于改进生成对抗网络的时序数据异常检测

针对时序数据复杂的时间相关性,以及现有异常检测模型存在准确性低、训练不稳定等问题,提出一种结合BiLSTM和WGAN-GP的无监督时序数据异常检测模型。使用BiLSTM作为生成器和判别器的基础网络来捕获时序数据的时间相关性;为保证训练过程的稳定性,使用Wasserstein距离取代原有的衡量方法,在判别器损失中加入梯度惩罚项;将重构损失与判别损失相结合定义异常函数,采用局部自适应阈值方法判别异常,提高异常检测的准确性。为验证模型性能,在涉及多个领域的5类数据集上进行实验,其结果表明,该模型相比于Arima、LSTM等模型具有最高的平均F1分数。

基于注意力特征融合时空图网络的超短期风电功率预测

为提高风电功率的预测精度,综合考虑时间和空间多维度因素的影响,提出了一种基于注意力机制和多阶段特征融合的时空图神经网络(Spatio-temporal graph neural network with attention mechanism and multistage feature fusion,AMF-STGNN)的超短期风电功率预测方法。首先基于Pearson相关系数法对数据特征进行降维,确定影响风电功率的关键因素。然后构建AMF-STGNN预测模型。该模型主要由时空关联网络构建模块和多维度时空特征抽取模块组成。通过时空关联网络构建模块构建时空图,以揭示风电气象因素的空间连接关系。通过多维度时空特征抽取模块应用时间卷积和图卷积挖掘数据的时空特征,并利用注意力机制学习重要特征。此外,该方法还引入残差结构和多阶段特征融合结构提高模型的表达能力。最后以宁夏3个风电场真实数据为例,验证了所提方法在提升风电功率预测精度方面的有效性。