A dynamic graph(DG)is adopted to portray the evolving interplay between nodes in real-world scenarios prevalently.A high-order graph convolutional network(HGCN)is equipped with the ability to represent a DG by the spa...A dynamic graph(DG)is adopted to portray the evolving interplay between nodes in real-world scenarios prevalently.A high-order graph convolutional network(HGCN)is equipped with the ability to represent a DG by the spatial-temporal message passing mechanism built on tensor product.Concretely,an HGCN utilizes the discrete Fourier transform(DFT)to implement temporal message passing and then employs face-wise product to realize spatial message passing.However,DFT is only a special case of assorted time-frequency transforms,which considers the complex temporal patterns partially,thereby resulting in an inaccurate temporal message passing possibly.To address this issue,this study proposes six advanced time-frequency transform-incorporated HGCNs(TF-HGCNs)with discrete Fourier,discrete Hartley,discrete cosine,Haar wavelet,Walsh Hadamard,and slant transforms.In addition,a potent ensemble is built regarding the proposed six TF-HGCNs as the bases.Finally,the corresponding theoretical proof is presented.Empirical studies on six DG datasets demonstrate that owing to diverse time-frequency transforms,the proposed six TF-HGCNs significantly outperform state-of-the-art models in addressing the task of link weight estimation.Moreover,their ensemble outstrips each base's performance.展开更多
This paper proposes a linear companding transform(CT)using either a single inflection point or two inflection points to reduce the peakto-average power ratio(PAPR)in orthogonal timefrequency space(OTFS)signals.The CT ...This paper proposes a linear companding transform(CT)using either a single inflection point or two inflection points to reduce the peakto-average power ratio(PAPR)in orthogonal timefrequency space(OTFS)signals.The CT strategically compresses higher amplitudes and enhances lower amplitudes based on carefully chosen scaling factors and points of inflection.With these selected parameters,the CT effectively reduces peak power while maintaining average power,leading to a substantial decrease in PAPR.We analyze noise changes in the inverse companding transform(ICT)process.The analysis reveals that the ICT amplifies less than 20%of the total noise.A convolutional encoder and soft decision Viterbi decoding algorithm are utilized in the OTFS system to improve the detection performance.We present simulation results focusing on PAPR reduction and bit error rate(BER)performance.These results demonstrate that the CT with two inflection points outperforms both the single inflection point case and the existingμ-law companding,clipping,peak windowing,unique OTFS frame structure,selected mapping,and partial transmit sequence methods,achieving significant PAPR reduction and BER performance.展开更多
Nonlinear transforms have significantly advanced learned image compression(LIC),particularly using residual blocks.This transform enhances the nonlinear expression ability and obtain compact feature representation by ...Nonlinear transforms have significantly advanced learned image compression(LIC),particularly using residual blocks.This transform enhances the nonlinear expression ability and obtain compact feature representation by enlarging the receptive field,which indicates how the convolution process extracts features in a high dimensional feature space.However,its functionality is restricted to the spatial dimension and network depth,limiting further improvements in network performance due to insufficient information interaction and representation.Crucially,the potential of high dimensional feature space in the channel dimension and the exploration of network width/resolution remain largely untapped.In this paper,we consider nonlinear transforms from the perspective of feature space,defining high-dimensional feature spaces in different dimensions and investigating the specific effects.Firstly,we introduce the dimension increasing and decreasing transforms in both channel and spatial dimensions to obtain high dimensional feature space and achieve better feature extraction.Secondly,we design a channel-spatial fusion residual transform(CSR),which incorporates multi-dimensional transforms for a more effective representation.Furthermore,we simplify the proposed fusion transform to obtain a slim architecture(CSR-sm),balancing network complexity and compression performance.Finally,we build the overall network with stacked CSR transforms to achieve better compression and reconstruction.Experimental results demonstrate that the proposed method can achieve superior ratedistortion performance compared to the existing LIC methods and traditional codecs.Specifically,our proposed method achieves 9.38%BD-rate reduction over VVC on Kodak dataset.展开更多
在新型电力系统复杂工况下,以策略表为主体、通过“离线仿真、在线匹配”的预案式频率稳定控制方案存在较高失配风险,甚至因调控失当引发二次冲击,严重威胁电力系统的安全稳定运行。提出一种计及预案式失配冲击的响应驱动频率稳定紧急...在新型电力系统复杂工况下,以策略表为主体、通过“离线仿真、在线匹配”的预案式频率稳定控制方案存在较高失配风险,甚至因调控失当引发二次冲击,严重威胁电力系统的安全稳定运行。提出一种计及预案式失配冲击的响应驱动频率稳定紧急切负荷策略。该策略动作在预案式控制之后,是对预案式控制的有益补充,能够有效提升系统频率稳定性。首先建立了基于系统频率响应(system frequency response,SFR)模型辨识的频率稳定切负荷量计算方法。提出了基于频率稀疏量测的SFR模型辨识方法,在此基础上建立了含稳定控制的SFR模型,根据频率稳定控制目标迭代求解切负荷量。其次,建立了基于Transformer网络的频率控制敏感点挖掘模型,通过分析关键发电机母线节点频率时序值和频率控制敏感点的映射关系,实现响应驱动的频率控制敏感点在线挖掘。最后,按照敏感点排序快速分配控制措施总量,构建频率稳定紧急控制方案。在某实际交直流混联万节点仿真系统验证了所提方法的有效性。展开更多
针对地图综合中建筑多边形化简方法依赖人工规则、自动化程度低且难以利用已有化简成果的问题,本文提出了一种基于Transformer机制的建筑多边形化简模型。该模型首先把建筑多边形映射至一定范围的网格空间,将建筑多边形的坐标串表达为...针对地图综合中建筑多边形化简方法依赖人工规则、自动化程度低且难以利用已有化简成果的问题,本文提出了一种基于Transformer机制的建筑多边形化简模型。该模型首先把建筑多边形映射至一定范围的网格空间,将建筑多边形的坐标串表达为网格序列,从而获取建筑多边形化简前后的Token序列,构建出建筑多边形化简样本对数据;随后采用Transformer架构建立模型,基于样本数据利用模型的掩码自注意力机制学习点序列之间的依赖关系,最终逐点生成新的简化多边形,从而实现建筑多边形的化简。在训练过程中,模型使用结构化的样本数据,设计了忽略特定索引的交叉熵损失函数以提升化简质量。试验设计包括主试验与泛化验证两部分。主试验基于洛杉矶1∶2000建筑数据集,分别采用0.2、0.3和0.5 mm 3种网格尺寸对多边形进行编码,实现了目标比例尺为1∶5000与1∶10000的化简。试验结果表明,在0.3 mm的网格尺寸下模型性能最优,验证集上的化简结果与人工标注的一致率超过92.0%,且针对北京部分区域的建筑多边形数据的泛化试验验证了模型的迁移能力;与LSTM模型的对比分析显示,在参数规模相近的条件下,LSTM模型无法形成有效收敛,并生成可用结果。本文证实了Transformer在处理空间几何序列任务中的潜力,且能够有效复用已有化简样本,为智能建筑多边形化简提供了具有工程实用价值的途径。展开更多
基金supported in part by the National Natural Science Foundation of China(62372385,62272078,62002337)Chongqing Natural Science Foundation(CSTB2022NSCQ-MSX1486,CSTB2023NSCQ-LZX0069)。
文摘A dynamic graph(DG)is adopted to portray the evolving interplay between nodes in real-world scenarios prevalently.A high-order graph convolutional network(HGCN)is equipped with the ability to represent a DG by the spatial-temporal message passing mechanism built on tensor product.Concretely,an HGCN utilizes the discrete Fourier transform(DFT)to implement temporal message passing and then employs face-wise product to realize spatial message passing.However,DFT is only a special case of assorted time-frequency transforms,which considers the complex temporal patterns partially,thereby resulting in an inaccurate temporal message passing possibly.To address this issue,this study proposes six advanced time-frequency transform-incorporated HGCNs(TF-HGCNs)with discrete Fourier,discrete Hartley,discrete cosine,Haar wavelet,Walsh Hadamard,and slant transforms.In addition,a potent ensemble is built regarding the proposed six TF-HGCNs as the bases.Finally,the corresponding theoretical proof is presented.Empirical studies on six DG datasets demonstrate that owing to diverse time-frequency transforms,the proposed six TF-HGCNs significantly outperform state-of-the-art models in addressing the task of link weight estimation.Moreover,their ensemble outstrips each base's performance.
文摘This paper proposes a linear companding transform(CT)using either a single inflection point or two inflection points to reduce the peakto-average power ratio(PAPR)in orthogonal timefrequency space(OTFS)signals.The CT strategically compresses higher amplitudes and enhances lower amplitudes based on carefully chosen scaling factors and points of inflection.With these selected parameters,the CT effectively reduces peak power while maintaining average power,leading to a substantial decrease in PAPR.We analyze noise changes in the inverse companding transform(ICT)process.The analysis reveals that the ICT amplifies less than 20%of the total noise.A convolutional encoder and soft decision Viterbi decoding algorithm are utilized in the OTFS system to improve the detection performance.We present simulation results focusing on PAPR reduction and bit error rate(BER)performance.These results demonstrate that the CT with two inflection points outperforms both the single inflection point case and the existingμ-law companding,clipping,peak windowing,unique OTFS frame structure,selected mapping,and partial transmit sequence methods,achieving significant PAPR reduction and BER performance.
基金supported by the Key Program of the National Natural Science Foundation of China(Grant No.62031013)Guangdong Province Key Construction Discipline Scientific Research Capacity Improvement Project(Grant No.2022ZDJS117).
文摘Nonlinear transforms have significantly advanced learned image compression(LIC),particularly using residual blocks.This transform enhances the nonlinear expression ability and obtain compact feature representation by enlarging the receptive field,which indicates how the convolution process extracts features in a high dimensional feature space.However,its functionality is restricted to the spatial dimension and network depth,limiting further improvements in network performance due to insufficient information interaction and representation.Crucially,the potential of high dimensional feature space in the channel dimension and the exploration of network width/resolution remain largely untapped.In this paper,we consider nonlinear transforms from the perspective of feature space,defining high-dimensional feature spaces in different dimensions and investigating the specific effects.Firstly,we introduce the dimension increasing and decreasing transforms in both channel and spatial dimensions to obtain high dimensional feature space and achieve better feature extraction.Secondly,we design a channel-spatial fusion residual transform(CSR),which incorporates multi-dimensional transforms for a more effective representation.Furthermore,we simplify the proposed fusion transform to obtain a slim architecture(CSR-sm),balancing network complexity and compression performance.Finally,we build the overall network with stacked CSR transforms to achieve better compression and reconstruction.Experimental results demonstrate that the proposed method can achieve superior ratedistortion performance compared to the existing LIC methods and traditional codecs.Specifically,our proposed method achieves 9.38%BD-rate reduction over VVC on Kodak dataset.
文摘在新型电力系统复杂工况下,以策略表为主体、通过“离线仿真、在线匹配”的预案式频率稳定控制方案存在较高失配风险,甚至因调控失当引发二次冲击,严重威胁电力系统的安全稳定运行。提出一种计及预案式失配冲击的响应驱动频率稳定紧急切负荷策略。该策略动作在预案式控制之后,是对预案式控制的有益补充,能够有效提升系统频率稳定性。首先建立了基于系统频率响应(system frequency response,SFR)模型辨识的频率稳定切负荷量计算方法。提出了基于频率稀疏量测的SFR模型辨识方法,在此基础上建立了含稳定控制的SFR模型,根据频率稳定控制目标迭代求解切负荷量。其次,建立了基于Transformer网络的频率控制敏感点挖掘模型,通过分析关键发电机母线节点频率时序值和频率控制敏感点的映射关系,实现响应驱动的频率控制敏感点在线挖掘。最后,按照敏感点排序快速分配控制措施总量,构建频率稳定紧急控制方案。在某实际交直流混联万节点仿真系统验证了所提方法的有效性。
文摘针对地图综合中建筑多边形化简方法依赖人工规则、自动化程度低且难以利用已有化简成果的问题,本文提出了一种基于Transformer机制的建筑多边形化简模型。该模型首先把建筑多边形映射至一定范围的网格空间,将建筑多边形的坐标串表达为网格序列,从而获取建筑多边形化简前后的Token序列,构建出建筑多边形化简样本对数据;随后采用Transformer架构建立模型,基于样本数据利用模型的掩码自注意力机制学习点序列之间的依赖关系,最终逐点生成新的简化多边形,从而实现建筑多边形的化简。在训练过程中,模型使用结构化的样本数据,设计了忽略特定索引的交叉熵损失函数以提升化简质量。试验设计包括主试验与泛化验证两部分。主试验基于洛杉矶1∶2000建筑数据集,分别采用0.2、0.3和0.5 mm 3种网格尺寸对多边形进行编码,实现了目标比例尺为1∶5000与1∶10000的化简。试验结果表明,在0.3 mm的网格尺寸下模型性能最优,验证集上的化简结果与人工标注的一致率超过92.0%,且针对北京部分区域的建筑多边形数据的泛化试验验证了模型的迁移能力;与LSTM模型的对比分析显示,在参数规模相近的条件下,LSTM模型无法形成有效收敛,并生成可用结果。本文证实了Transformer在处理空间几何序列任务中的潜力,且能够有效复用已有化简样本,为智能建筑多边形化简提供了具有工程实用价值的途径。
