This study proposes a novel radar echo extrapolation algorithm,OF-ConvGRU,which integrates Optical Flow(OF)and Convolutional Gated Recurrent Unit(ConvGRU)methods for improved nowcasting.Using the Standardized Radar Da...This study proposes a novel radar echo extrapolation algorithm,OF-ConvGRU,which integrates Optical Flow(OF)and Convolutional Gated Recurrent Unit(ConvGRU)methods for improved nowcasting.Using the Standardized Radar Dataset of the Guangdong-Hong Kong-Macao Greater Bay Area,the performance of OF-ConvGRU was evaluated against OF and ConvGRU methods.Threat Score(TS)and Bias Score(BIAS)were employed to assess extrapolation accuracy across various echo intensities(20-50 dBz)and weather phenomena.Results demonstrate that OF-ConvGRU significantly enhances prediction accuracy for moderate-intensity echoes(30-40 dBz),effectively combining OF s precise motion estimation with ConvGRU s nonlinear learning capabilities.However,challenges persist in low-intensity(20 dBz)and high-intensity(50 dBz)echo predictions.The study reveals distinct advantages of each method in specific contexts,highlighting the importance of multi-method approaches in operational nowcasting.OF-ConvGRU shows promise in balancing short-term accuracy with long-term stability,particularly for complex weather systems.展开更多
Extrapolation on Temporal Knowledge Graphs(TKGs)aims to predict future knowledge from a set of historical Knowledge Graphs in chronological order.The temporally adjacent facts in TKGs naturally form event sequences,ca...Extrapolation on Temporal Knowledge Graphs(TKGs)aims to predict future knowledge from a set of historical Knowledge Graphs in chronological order.The temporally adjacent facts in TKGs naturally form event sequences,called event evolution patterns,implying informative temporal dependencies between events.Recently,many extrapolation works on TKGs have been devoted to modelling these evolutional patterns,but the task is still far from resolved because most existing works simply rely on encoding these patterns into entity representations while overlooking the significant information implied by relations of evolutional patterns.However,the authors realise that the temporal dependencies inherent in the relations of these event evolution patterns may guide the follow-up event prediction to some extent.To this end,a Temporal Relational Context-based Temporal Dependencies Learning Network(TRenD)is proposed to explore the temporal context of relations for more comprehensive learning of event evolution patterns,especially those temporal dependencies caused by interactive patterns of relations.Trend incorporates a semantic context unit to capture semantic correlations between relations,and a structural context unit to learn the interaction pattern of relations.By learning the temporal contexts of relations semantically and structurally,the authors gain insights into the underlying event evolution patterns,enabling to extract comprehensive historical information for future prediction better.Experimental results on benchmark datasets demonstrate the superiority of the model.展开更多
离线强化学习旨在仅通过使用预先收集的离线数据集进行策略的有效学习,从而减少与环境直接交互所带来的高昂成本。然而,由于缺少环境对智能体行为的交互反馈,从离线数据集中学习到的策略可能会遇到数据分布偏移的问题,进而导致外推误差...离线强化学习旨在仅通过使用预先收集的离线数据集进行策略的有效学习,从而减少与环境直接交互所带来的高昂成本。然而,由于缺少环境对智能体行为的交互反馈,从离线数据集中学习到的策略可能会遇到数据分布偏移的问题,进而导致外推误差的不断加剧。当前方法多采用策略约束或模仿学习方法来缓解这一问题,但其学习到的策略通常较为保守。针对上述难题,提出一种基于自适应分位数的方法。具体而言,该方法在双Q估计的基础上进一步利用双Q的估计差值大小对分布外未知动作的价值高估情况进行评估,同时结合分位数思想自适应调整分位数来校正过估计偏差。此外,构建分位数优势函数作为策略约束项权重以平衡智能体对数据集的探索和模仿,从而缓解策略学习的保守性。最后在D4RL(datasets for deep data-driven reinforcement learning)数据集上验证算法的有效性,该算法在多个任务数据集上表现优异,同时展现出在不同场景应用下的广泛潜力。展开更多
随着科学技术的不断进步、替代毒理学的快速发展以及动物实验禁令在部分国家和地区的推广,传统风险评估的局限性日益凸显,下一代风险评估(next generation risk assessment,NGRA)应运而生。NGRA是一种替代动物实验的新型风险评估方法,...随着科学技术的不断进步、替代毒理学的快速发展以及动物实验禁令在部分国家和地区的推广,传统风险评估的局限性日益凸显,下一代风险评估(next generation risk assessment,NGRA)应运而生。NGRA是一种替代动物实验的新型风险评估方法,它依赖于通过体外测试、体外-体内外推(in vitro to in vivo extrapolation,IVIVE)、计算毒理学、交叉参照等新路线方法(new approach methodologies,NAMs)生成的数据,这些方法使用基于人类的模型,准确地反映了人类生物学,增加了风险评估的准确性以及高效性。本文系统整理了NGRA的研究现状和进展,简要介绍了NGRA的框架,主要围绕NGRA采用的NAMs及面临的挑战进行了重点分析,同时分享多种暴露场景下的应用案例,并对NGRA未来研究方向进行展望,以期为我国化学物质环境管理提供更好的方法学支撑。展开更多
Autoregressive (AR) modeling is applied to data extrapolation of radio frequency (RF) echo signals, and Burg algorithm, which can be computed in small amount and lead to a stable prediction filter, is used to estimate...Autoregressive (AR) modeling is applied to data extrapolation of radio frequency (RF) echo signals, and Burg algorithm, which can be computed in small amount and lead to a stable prediction filter, is used to estimate the prediction parameters of AR modeling. The complex data samples are directly extrapolated to obtain the extrapolated echo data in the frequency domain. The small rotating angle data extrapolation and the large rotating angular data extrapolation are considered separately in azimuth domain. The method of data extrapolation for the small rotating angle is the same as that in frequency domain, while the amplitude samples of large rotating angle echo data are extrapolated to obtain extrapolated echo amplitude, and the complex data of large rotating angle echo samples are extrapolated to get the extrapolated echo phase respectively. The calculation results show that the extrapolated echo data obtained by the above mentioned methods are accurate.展开更多
基金Scientific Research and Development Project of Hebei Meteorological Bureau(23ky08).
文摘This study proposes a novel radar echo extrapolation algorithm,OF-ConvGRU,which integrates Optical Flow(OF)and Convolutional Gated Recurrent Unit(ConvGRU)methods for improved nowcasting.Using the Standardized Radar Dataset of the Guangdong-Hong Kong-Macao Greater Bay Area,the performance of OF-ConvGRU was evaluated against OF and ConvGRU methods.Threat Score(TS)and Bias Score(BIAS)were employed to assess extrapolation accuracy across various echo intensities(20-50 dBz)and weather phenomena.Results demonstrate that OF-ConvGRU significantly enhances prediction accuracy for moderate-intensity echoes(30-40 dBz),effectively combining OF s precise motion estimation with ConvGRU s nonlinear learning capabilities.However,challenges persist in low-intensity(20 dBz)and high-intensity(50 dBz)echo predictions.The study reveals distinct advantages of each method in specific contexts,highlighting the importance of multi-method approaches in operational nowcasting.OF-ConvGRU shows promise in balancing short-term accuracy with long-term stability,particularly for complex weather systems.
基金supported in part by the National Natural Science Foundation of China(No.62302507)and the funding of Harbin Institute of Technology(Shenzhen)(No.20210035).
文摘Extrapolation on Temporal Knowledge Graphs(TKGs)aims to predict future knowledge from a set of historical Knowledge Graphs in chronological order.The temporally adjacent facts in TKGs naturally form event sequences,called event evolution patterns,implying informative temporal dependencies between events.Recently,many extrapolation works on TKGs have been devoted to modelling these evolutional patterns,but the task is still far from resolved because most existing works simply rely on encoding these patterns into entity representations while overlooking the significant information implied by relations of evolutional patterns.However,the authors realise that the temporal dependencies inherent in the relations of these event evolution patterns may guide the follow-up event prediction to some extent.To this end,a Temporal Relational Context-based Temporal Dependencies Learning Network(TRenD)is proposed to explore the temporal context of relations for more comprehensive learning of event evolution patterns,especially those temporal dependencies caused by interactive patterns of relations.Trend incorporates a semantic context unit to capture semantic correlations between relations,and a structural context unit to learn the interaction pattern of relations.By learning the temporal contexts of relations semantically and structurally,the authors gain insights into the underlying event evolution patterns,enabling to extract comprehensive historical information for future prediction better.Experimental results on benchmark datasets demonstrate the superiority of the model.
文摘离线强化学习旨在仅通过使用预先收集的离线数据集进行策略的有效学习,从而减少与环境直接交互所带来的高昂成本。然而,由于缺少环境对智能体行为的交互反馈,从离线数据集中学习到的策略可能会遇到数据分布偏移的问题,进而导致外推误差的不断加剧。当前方法多采用策略约束或模仿学习方法来缓解这一问题,但其学习到的策略通常较为保守。针对上述难题,提出一种基于自适应分位数的方法。具体而言,该方法在双Q估计的基础上进一步利用双Q的估计差值大小对分布外未知动作的价值高估情况进行评估,同时结合分位数思想自适应调整分位数来校正过估计偏差。此外,构建分位数优势函数作为策略约束项权重以平衡智能体对数据集的探索和模仿,从而缓解策略学习的保守性。最后在D4RL(datasets for deep data-driven reinforcement learning)数据集上验证算法的有效性,该算法在多个任务数据集上表现优异,同时展现出在不同场景应用下的广泛潜力。
文摘随着科学技术的不断进步、替代毒理学的快速发展以及动物实验禁令在部分国家和地区的推广,传统风险评估的局限性日益凸显,下一代风险评估(next generation risk assessment,NGRA)应运而生。NGRA是一种替代动物实验的新型风险评估方法,它依赖于通过体外测试、体外-体内外推(in vitro to in vivo extrapolation,IVIVE)、计算毒理学、交叉参照等新路线方法(new approach methodologies,NAMs)生成的数据,这些方法使用基于人类的模型,准确地反映了人类生物学,增加了风险评估的准确性以及高效性。本文系统整理了NGRA的研究现状和进展,简要介绍了NGRA的框架,主要围绕NGRA采用的NAMs及面临的挑战进行了重点分析,同时分享多种暴露场景下的应用案例,并对NGRA未来研究方向进行展望,以期为我国化学物质环境管理提供更好的方法学支撑。
文摘Autoregressive (AR) modeling is applied to data extrapolation of radio frequency (RF) echo signals, and Burg algorithm, which can be computed in small amount and lead to a stable prediction filter, is used to estimate the prediction parameters of AR modeling. The complex data samples are directly extrapolated to obtain the extrapolated echo data in the frequency domain. The small rotating angle data extrapolation and the large rotating angular data extrapolation are considered separately in azimuth domain. The method of data extrapolation for the small rotating angle is the same as that in frequency domain, while the amplitude samples of large rotating angle echo data are extrapolated to obtain extrapolated echo amplitude, and the complex data of large rotating angle echo samples are extrapolated to get the extrapolated echo phase respectively. The calculation results show that the extrapolated echo data obtained by the above mentioned methods are accurate.
