Predicting potential facts in the future,Temporal Knowledge Graph(TKG)extrapolation remains challenging because of the deep dependence between the temporal association and semantic patterns of facts.Intuitively,facts(...Predicting potential facts in the future,Temporal Knowledge Graph(TKG)extrapolation remains challenging because of the deep dependence between the temporal association and semantic patterns of facts.Intuitively,facts(events)that happened at different timestamps have different influences on future events,which can be attributed to a hierarchy among not only facts but also relevant entities.Therefore,it is crucial to pay more attention to important entities and events when forecasting the future.However,most existing methods focus on reasoning over temporally evolving facts or mining evolutional patterns from known facts,which may be affected by the diversity and variability of the evolution,and they might fail to attach importance to facts that matter.Hyperbolic geometry was proved to be effective in capturing hierarchical patterns among data,which is considered to be a solution for modelling hierarchical relations among facts.To this end,we propose ReTIN,a novel model integrating real-time influence of historical facts for TKG reasoning based on hyperbolic geometry,which provides low-dimensional embeddings to capture latent hierarchical structures and other rich semantic patterns of the existing TKG.Considering both real-time and global features of TKG boosts the adaptation of ReTIN to the ever-changing dynamics and inherent constraints.Extensive experiments on benchmarks demonstrate the superiority of ReTIN over various baselines.The ablation study further supports the value of exploiting temporal information.展开更多
Accurate precipitation nowcasting can provide great convenience to the public so they can conduct corresponding arrangements in advance to deal with the possible impact of upcoming heavy rain.Recent relevant research ...Accurate precipitation nowcasting can provide great convenience to the public so they can conduct corresponding arrangements in advance to deal with the possible impact of upcoming heavy rain.Recent relevant research activities have shown their concerns on various deep learning models for radar echo extrapolation,where radar echo maps were used to predict their consequent moment,so as to recognize potential severe convective weather events.However,these approaches suffer from an inaccurate prediction of echo dynamics and unreliable depiction of echo aggregation or dissipation,due to the size limitation of convolution filter,lack of global feature,and less attention to features from previous states.To address the problems,this paper proposes a CEMA-LSTM recurrent unit,which is embedded with a Contextual Feature Correlation Enhancement Block(CEB)and a Multi-Attention Mechanism Block(MAB).The CEB enhances contextual feature correlation and supports its model to memorize significant features for near-future prediction;the MAB uses a position and channel attention mechanism to capture global features of radar echoes.Two practical radar echo datasets were used involving the FREM and CIKM 2017 datasets.Both quantification and visualization of comparative experimental results have demonstrated outperformance of the proposed CEMA-LSTMover recentmodels,e.g.,PhyDNet,MIM and PredRNN++,etc.In particular,compared with the second-rankedmodel,its average POD,FAR and CSI have been improved by 3.87%,1.65%and 1.79%,respectively on the FREM,and by 1.42%,5.60%and 3.16%,respectively on the CIKM 2017.展开更多
Accurate measurements of upwelling irradiance just beneath the ocean surface,E_(u)(λ,0^(-)),can be used to calculate ocean optical parameters,and further develop retrieval algorithms for remotely sensing water compon...Accurate measurements of upwelling irradiance just beneath the ocean surface,E_(u)(λ,0^(-)),can be used to calculate ocean optical parameters,and further develop retrieval algorithms for remotely sensing water component concentrations.Due to the effects of sea surface waves,perturbation from instrument platform(ship),and instrument self-shading,E_(u)(λ,0^(-))is often difficult to be accurately measured.This study presents a procedure for extrapolating the E_(u)(λ,0^(-))from the in-water radiometric profile measurements.Using the optical profile data from 13 bands(ranging from 381 to 779 nm)measured by 45 casts in the Ligurian Sea during 2003–2009,the E_(u)(λ,0^(-))was extrapolated from in-water upwelling irradiance measurements between the initial shallow depth,Z_(0),and an optimal bottom depth,Z_(1),by three linear models(linear,2-degree polynomial,and exponential)and two nonlinear models(LOESS and spline).The accumulated errors of extrapolated E_(u)(λ,0^(-))at each wavelength for the five models were calculated.It was found that the optimal Z_(1) depth for the linear and exponential models was at the depth of80%of E_(u)(λ,Z_(0)),50%of E_(u)(λ,Z_(0))for the 2-degree polynomial model,40%of E_(u)(λ,Z_(0))for the LOESS model,and 15%of E_(u)(λ,Z_(0))for the spline model.The extrapolated E_(u)(λ,0^(-))derived from the five models was in good agreement with the calculated true E_(u)(λ,0^(-)).In all bands,the 2-degree polynomial model achieved the highest accuracy,followed by the LOESS model.In the short band of 381–559 nm,the linear and exponential models had the third-best performance,and the spline model performed worst within this range.For the red band of 619–779 nm,the accuracies of the exponential and spline models had the third highest performance,and the linear model produced lowest accuracy.Hence,the 2-degree polynomial model was an optimal procedure for extrapolation of E_(u)(λ,0^(-))from the in-water radiometric profile measurements.展开更多
Weather radar echo extrapolation plays a crucial role in weather forecasting.However,traditional weather radar echo extrapolation methods are not very accurate and do not make full use of historical data.Deep learning...Weather radar echo extrapolation plays a crucial role in weather forecasting.However,traditional weather radar echo extrapolation methods are not very accurate and do not make full use of historical data.Deep learning algorithms based on Recurrent Neural Networks also have the problem of accumulating errors.Moreover,it is difficult to obtain higher accuracy by relying on a single historical radar echo observation.Therefore,in this study,we constructed the Fusion GRU module,which leverages a cascade structure to effectively combine radar echo data and mean wind data.We also designed the Top Connection so that the model can capture the global spatial relationship to construct constraints on the predictions.Based on the Jiangsu Province dataset,we compared some models.The results show that our proposed model,Cascade Fusion Spatiotemporal Network(CFSN),improved the critical success index(CSI)by 10.7%over the baseline at the threshold of 30 dBZ.Ablation experiments further validated the effectiveness of our model.Similarly,the CSI of the complete CFSN was 0.004 higher than the suboptimal solution without the cross-attention module at the threshold of 30 dBZ.展开更多
In this paper, a new extrapolation economy cascadic multigrid method is proposed to solve the image restoration model. The new method combines the new extrapolation formula and quadratic interpolation to design a nonl...In this paper, a new extrapolation economy cascadic multigrid method is proposed to solve the image restoration model. The new method combines the new extrapolation formula and quadratic interpolation to design a nonlinear prolongation operator, which provides more accurate initial values for the fine grid level. An edge preserving denoising operator is constructed to remove noise and preserve image edges. The local smoothing operator reduces the influence of staircase effect. The experiment results show that the new method not only improves the computational efficiency but also ensures good recovery quality.展开更多
基金Major Key Project of Pengcheng Laboratory,Grant/Award Number:PCL2022A03。
文摘Predicting potential facts in the future,Temporal Knowledge Graph(TKG)extrapolation remains challenging because of the deep dependence between the temporal association and semantic patterns of facts.Intuitively,facts(events)that happened at different timestamps have different influences on future events,which can be attributed to a hierarchy among not only facts but also relevant entities.Therefore,it is crucial to pay more attention to important entities and events when forecasting the future.However,most existing methods focus on reasoning over temporally evolving facts or mining evolutional patterns from known facts,which may be affected by the diversity and variability of the evolution,and they might fail to attach importance to facts that matter.Hyperbolic geometry was proved to be effective in capturing hierarchical patterns among data,which is considered to be a solution for modelling hierarchical relations among facts.To this end,we propose ReTIN,a novel model integrating real-time influence of historical facts for TKG reasoning based on hyperbolic geometry,which provides low-dimensional embeddings to capture latent hierarchical structures and other rich semantic patterns of the existing TKG.Considering both real-time and global features of TKG boosts the adaptation of ReTIN to the ever-changing dynamics and inherent constraints.Extensive experiments on benchmarks demonstrate the superiority of ReTIN over various baselines.The ablation study further supports the value of exploiting temporal information.
基金funding from the Key Laboratory Foundation of National Defence Technology under Grant 61424010208National Natural Science Foundation of China(Nos.62002276,41911530242 and 41975142)+3 种基金5150 Spring Specialists(05492018012 and 05762018039)Major Program of the National Social Science Fund of China(Grant No.17ZDA092)333 High-LevelTalent Cultivation Project of Jiangsu Province(BRA2018332)Royal Society of Edinburgh,UK andChina Natural Science Foundation Council(RSE Reference:62967)_Liu)_2018)_2)under their Joint International Projects Funding Scheme and Basic Research Programs(Natural Science Foundation)of Jiangsu Province(BK20191398 and BK20180794).
文摘Accurate precipitation nowcasting can provide great convenience to the public so they can conduct corresponding arrangements in advance to deal with the possible impact of upcoming heavy rain.Recent relevant research activities have shown their concerns on various deep learning models for radar echo extrapolation,where radar echo maps were used to predict their consequent moment,so as to recognize potential severe convective weather events.However,these approaches suffer from an inaccurate prediction of echo dynamics and unreliable depiction of echo aggregation or dissipation,due to the size limitation of convolution filter,lack of global feature,and less attention to features from previous states.To address the problems,this paper proposes a CEMA-LSTM recurrent unit,which is embedded with a Contextual Feature Correlation Enhancement Block(CEB)and a Multi-Attention Mechanism Block(MAB).The CEB enhances contextual feature correlation and supports its model to memorize significant features for near-future prediction;the MAB uses a position and channel attention mechanism to capture global features of radar echoes.Two practical radar echo datasets were used involving the FREM and CIKM 2017 datasets.Both quantification and visualization of comparative experimental results have demonstrated outperformance of the proposed CEMA-LSTMover recentmodels,e.g.,PhyDNet,MIM and PredRNN++,etc.In particular,compared with the second-rankedmodel,its average POD,FAR and CSI have been improved by 3.87%,1.65%and 1.79%,respectively on the FREM,and by 1.42%,5.60%and 3.16%,respectively on the CIKM 2017.
基金Supported by the Marine Special Program of Jiangsu Province in China (No.JSZRHYKJ202007)the National Natural Science Foundation of China (No.40801145)。
文摘Accurate measurements of upwelling irradiance just beneath the ocean surface,E_(u)(λ,0^(-)),can be used to calculate ocean optical parameters,and further develop retrieval algorithms for remotely sensing water component concentrations.Due to the effects of sea surface waves,perturbation from instrument platform(ship),and instrument self-shading,E_(u)(λ,0^(-))is often difficult to be accurately measured.This study presents a procedure for extrapolating the E_(u)(λ,0^(-))from the in-water radiometric profile measurements.Using the optical profile data from 13 bands(ranging from 381 to 779 nm)measured by 45 casts in the Ligurian Sea during 2003–2009,the E_(u)(λ,0^(-))was extrapolated from in-water upwelling irradiance measurements between the initial shallow depth,Z_(0),and an optimal bottom depth,Z_(1),by three linear models(linear,2-degree polynomial,and exponential)and two nonlinear models(LOESS and spline).The accumulated errors of extrapolated E_(u)(λ,0^(-))at each wavelength for the five models were calculated.It was found that the optimal Z_(1) depth for the linear and exponential models was at the depth of80%of E_(u)(λ,Z_(0)),50%of E_(u)(λ,Z_(0))for the 2-degree polynomial model,40%of E_(u)(λ,Z_(0))for the LOESS model,and 15%of E_(u)(λ,Z_(0))for the spline model.The extrapolated E_(u)(λ,0^(-))derived from the five models was in good agreement with the calculated true E_(u)(λ,0^(-)).In all bands,the 2-degree polynomial model achieved the highest accuracy,followed by the LOESS model.In the short band of 381–559 nm,the linear and exponential models had the third-best performance,and the spline model performed worst within this range.For the red band of 619–779 nm,the accuracies of the exponential and spline models had the third highest performance,and the linear model produced lowest accuracy.Hence,the 2-degree polynomial model was an optimal procedure for extrapolation of E_(u)(λ,0^(-))from the in-water radiometric profile measurements.
基金National Natural Science Foundation of China(42375145)The Open Grants of China Meteorological Admin-istration Radar Meteorology Key Laboratory(2023LRM-A02)。
文摘Weather radar echo extrapolation plays a crucial role in weather forecasting.However,traditional weather radar echo extrapolation methods are not very accurate and do not make full use of historical data.Deep learning algorithms based on Recurrent Neural Networks also have the problem of accumulating errors.Moreover,it is difficult to obtain higher accuracy by relying on a single historical radar echo observation.Therefore,in this study,we constructed the Fusion GRU module,which leverages a cascade structure to effectively combine radar echo data and mean wind data.We also designed the Top Connection so that the model can capture the global spatial relationship to construct constraints on the predictions.Based on the Jiangsu Province dataset,we compared some models.The results show that our proposed model,Cascade Fusion Spatiotemporal Network(CFSN),improved the critical success index(CSI)by 10.7%over the baseline at the threshold of 30 dBZ.Ablation experiments further validated the effectiveness of our model.Similarly,the CSI of the complete CFSN was 0.004 higher than the suboptimal solution without the cross-attention module at the threshold of 30 dBZ.
文摘In this paper, a new extrapolation economy cascadic multigrid method is proposed to solve the image restoration model. The new method combines the new extrapolation formula and quadratic interpolation to design a nonlinear prolongation operator, which provides more accurate initial values for the fine grid level. An edge preserving denoising operator is constructed to remove noise and preserve image edges. The local smoothing operator reduces the influence of staircase effect. The experiment results show that the new method not only improves the computational efficiency but also ensures good recovery quality.
