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
The Besov spaces B_p^(α,4)(Γ)and Triebel-Lizorkin spaces F_p^(α,4)(Γ)with high order x∈R on a Lipschitz curve Γ are defind,when 1≤p≤∞,1≤q≤∞.To compare to the classical case.a difference characterization of...The Besov spaces B_p^(α,4)(Γ)and Triebel-Lizorkin spaces F_p^(α,4)(Γ)with high order x∈R on a Lipschitz curve Γ are defind,when 1≤p≤∞,1≤q≤∞.To compare to the classical case.a difference characterization of such spaces in the case|x|<1 is given also.展开更多
High concentration of population and other elements brought by high urbanization has pushed the future construction of urban public spaces to be both "compact" and "ecological". Theory and methodol...High concentration of population and other elements brought by high urbanization has pushed the future construction of urban public spaces to be both "compact" and "ecological". Theory and methodology of landscape urbanism provides a new perspective for the construction of public spaces in high-density urban environment. Based on the concept and basic standpoints of landscape urbanism, this paper elaborated its potential value in spatial connection, functional integration and vitality promotion, and also analyzed practical strategies of landscape urbanism from three perspectives of structure, fold and stage by combining with typical cases, so as to provide references for designing urban public spaces in high-density environment.展开更多
[Objective] The paper was to understand the physiological response of space flight mutation new strains of Festuca arundinacea to high temperature stress. [Method] The influence of high temperature stress on eco-physi...[Objective] The paper was to understand the physiological response of space flight mutation new strains of Festuca arundinacea to high temperature stress. [Method] The influence of high temperature stress on eco-physiological characteristics of 11 F. arundinacea materials was studied in a pot experiment.Physiological and biochemical indexes,including soluble protein content,superoxide dismutase( SOD),peroxidase( POD) and catalase( CAT) activities,were measured and analyzed once every 4 d for a total of three times. Subordinate function and grey rational analysis were used to comprehensively evaluate the heat resistance of 11 materials. [Result] The soluble protein content decreased with the extension of heat resistance stress,while SOD,POD and CAT activities showed an upward trend. The high temperature resistance of 11 F. arundinacea materials was evaluated by subordinate function method. The resistance order was SP5-85> SP5-60 > SP5-71 > SP5-42 > SP5-94 > SP5-7 > SP5-88 > F. arundinacea cv Shuicheng > SP5-5 > SP5-89 = F. arundinacea cv Qiancao No. 1. The relational order of various heat resistance indexes and heat resistance obtained by grey rational analysis was SOD > CAT > POD > soluble protein content. [Conclusion]The result provides a theoretical basis for stress physiology and stress breeding of cold season grass in southwest region.展开更多
In this paper, we propose a new differential space-time-frequency （DSTF） modulation for MIMOOFDM system with four transmit-antennas and arbitrary receive-antennas, which can improve the transmission rate since it ca...In this paper, we propose a new differential space-time-frequency （DSTF） modulation for MIMOOFDM system with four transmit-antennas and arbitrary receive-antennas, which can improve the transmission rate since it can adopt high order quadrature amplitude modulation （QAM） modulation. Our proposed DSTF scheme embeds some full diversity full rate （FDFR） quasi-orthogonal space-time codes （QOSTBC） with QAM modulation into the frequency intervals and adopts the differential modulation in both time and frequency domains. The simulation results demonstrate that the proposed DSTF scheme can improve transmission rate greatly. Compared with the conventional differential unitary space-time modulation （DUSTM）, it can get better transmission performance in high transmission rate for MIMO-OFDM system.展开更多
As the most influential trade platform for yarn and fiber industry players in Asia,Yarn Expo Autumn is renowned for its comprehensive range of exhibitors.As such,the show keeps gaining in popularity,with the number of...As the most influential trade platform for yarn and fiber industry players in Asia,Yarn Expo Autumn is renowned for its comprehensive range of exhibitors.As such,the show keeps gaining in popularity,with the number of exhibitors increasing from 258 to morethan 270 this year.To date,they come from 12 countries展开更多
利用CMA-MESO和SWC-WARMS高分辨率模式2023年5—9月小时降水预报产品,四川自动站小时降水资料及CLDAS(CMA Land Data Assimilation System)三源融合小时降水资料,采用“点对点”和“点对面”检验方法,对两家高分辨率模式小时降水产品在...利用CMA-MESO和SWC-WARMS高分辨率模式2023年5—9月小时降水预报产品,四川自动站小时降水资料及CLDAS(CMA Land Data Assimilation System)三源融合小时降水资料,采用“点对点”和“点对面”检验方法,对两家高分辨率模式小时降水产品在四川地区的预报性能进行评估。主要结论如下:①小时降水10 mm以下和50 mm以上的极端降水,CMA-MESO的预报参考性显著优于SWC-WARMS;小时降水10~30 mm,5—6月以SWC-WARMS表现更优,7—9月则以CMA-MESO表现更佳。②空间偏差特征分析表明,两家模式的小时降水平均绝对误差空间分布整体较为一致,在盆周山区、川西高原及凉山州北部误差较大,模式误差和实况降水强度、海拔高度呈一定正相关,且SWC-WARMS的空间误差更大。③时间偏差特征分析表明,CMA-MESO的短时强降水站点频次峰值时间偏差为1~2 h,而SWC-WARMS达3~4 h。④为优化TS评分,针对CMA-MESO和SWC-WARMS模式,在空间上,预报时可分别考虑邻域半径9 km和12 km内出现对应量级降水的可能性;在时间上,对于5~30 mm小时降水,预报时可考虑预报时刻前后1 h和2 h发生的可能性,其他量级可考虑前后1 h。展开更多
随着大语言模型(large language models,LLMs)(以下简称“大模型”)参数规模的持续增长,微调百亿级参数大模型对计算和存储资源提出了极高要求。传统分布式训练方案通常依赖大量高端GPU和高速互联网络,训练成本极为昂贵。现有单GPU训练...随着大语言模型(large language models,LLMs)(以下简称“大模型”)参数规模的持续增长,微调百亿级参数大模型对计算和存储资源提出了极高要求。传统分布式训练方案通常依赖大量高端GPU和高速互联网络,训练成本极为昂贵。现有单GPU训练方案虽通过张量卸载缓解显存压力,但仍然面临I/O传输效率低和设备利用率不足等问题。传统内核态I/O操作在大规模张量迁移中引入频繁的系统调用和上下文切换,成为制约性能的关键瓶颈;同时,优化器计算无法充分发挥多核CPU的并行能力,难以实现与GPU计算的有效重叠,进一步限制了系统性能。针对上述问题,提出了一种面向大模型训练的异构内存卸载与I/O优化方案HiTrain。首先构建了基于存储性能开发工具包(storage performance development kit,SPDK)的高性能张量存储模块,通过在用户态管理张量数据,避免了内核I/O栈开销,从而提高张量卸载的并发性与吞吐率;其次,设计并实现了基于异步优化器的存储-计算流水线调度模块,通过对优化器的执行进行优化重排来减少GPU等待时间,提高整体训练效率。实验结果表明,在配备单张GPU和非易失性存储器快速固态硬盘(non-volatile memory express solid state drive,NVMe SSD)的服务器上,所提出的方案能够充分利用系统中的存算资源,使得模型训练过程中张量卸载与加载效率提升32.7%,整体训练吞吐提升至现有方案的1.49倍,为低成本大模型训练提供了切实可行的技术路径。展开更多
随机开关频率调制策略通过将开关频率在扩频范围内随机变化抑制高频谐波。虽然过宽的扩频范围能更有效地抑制高频谐波,但同时会带来严重的电流畸变和较大的转矩脉动。为此,该文提出在窄扩频范围下考虑随机数变化的随机开关频率空间矢量...随机开关频率调制策略通过将开关频率在扩频范围内随机变化抑制高频谐波。虽然过宽的扩频范围能更有效地抑制高频谐波,但同时会带来严重的电流畸变和较大的转矩脉动。为此,该文提出在窄扩频范围下考虑随机数变化的随机开关频率空间矢量脉宽调制(random switching frequency space vector pulse width modulation,RSF-SVPWM)。与传统的RSF-SVPWM相比,提出的方法通过限制相邻随机数变化的最大值,减小了相邻随机数的相对变化,使高频谐波由三角形分布变为梯形分布,从而降低了高频谐波幅值,实现窄扩频范围内对高频谐波的有效抑制。实验结果表明,在相同的扩频范围内,该方法更有效地降低了开关频率及其整数倍处谐波的峰值。展开更多
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.
基金The author is supported in part by the Foundation of Zhongshan University Advanced Research Centre and NSF of China.
文摘The Besov spaces B_p^(α,4)(Γ)and Triebel-Lizorkin spaces F_p^(α,4)(Γ)with high order x∈R on a Lipschitz curve Γ are defind,when 1≤p≤∞,1≤q≤∞.To compare to the classical case.a difference characterization of such spaces in the case|x|<1 is given also.
文摘High concentration of population and other elements brought by high urbanization has pushed the future construction of urban public spaces to be both "compact" and "ecological". Theory and methodology of landscape urbanism provides a new perspective for the construction of public spaces in high-density urban environment. Based on the concept and basic standpoints of landscape urbanism, this paper elaborated its potential value in spatial connection, functional integration and vitality promotion, and also analyzed practical strategies of landscape urbanism from three perspectives of structure, fold and stage by combining with typical cases, so as to provide references for designing urban public spaces in high-density environment.
基金Supported by United Fund of Guizhou Province(QKH J LKN[2013]24)Key Special Project of Guizhou Province(QKHZDZX[2014]6017)High-level Innovative Talents Training of Guizhou Province(QKHRC[2016]4024)
文摘[Objective] The paper was to understand the physiological response of space flight mutation new strains of Festuca arundinacea to high temperature stress. [Method] The influence of high temperature stress on eco-physiological characteristics of 11 F. arundinacea materials was studied in a pot experiment.Physiological and biochemical indexes,including soluble protein content,superoxide dismutase( SOD),peroxidase( POD) and catalase( CAT) activities,were measured and analyzed once every 4 d for a total of three times. Subordinate function and grey rational analysis were used to comprehensively evaluate the heat resistance of 11 materials. [Result] The soluble protein content decreased with the extension of heat resistance stress,while SOD,POD and CAT activities showed an upward trend. The high temperature resistance of 11 F. arundinacea materials was evaluated by subordinate function method. The resistance order was SP5-85> SP5-60 > SP5-71 > SP5-42 > SP5-94 > SP5-7 > SP5-88 > F. arundinacea cv Shuicheng > SP5-5 > SP5-89 = F. arundinacea cv Qiancao No. 1. The relational order of various heat resistance indexes and heat resistance obtained by grey rational analysis was SOD > CAT > POD > soluble protein content. [Conclusion]The result provides a theoretical basis for stress physiology and stress breeding of cold season grass in southwest region.
基金This work was supported in part by the National Natural Science Foundation of China under grant No. 60572117the Natural Science Foundation of Hubei Province under grant No. 2005ABA244.
文摘In this paper, we propose a new differential space-time-frequency （DSTF） modulation for MIMOOFDM system with four transmit-antennas and arbitrary receive-antennas, which can improve the transmission rate since it can adopt high order quadrature amplitude modulation （QAM） modulation. Our proposed DSTF scheme embeds some full diversity full rate （FDFR） quasi-orthogonal space-time codes （QOSTBC） with QAM modulation into the frequency intervals and adopts the differential modulation in both time and frequency domains. The simulation results demonstrate that the proposed DSTF scheme can improve transmission rate greatly. Compared with the conventional differential unitary space-time modulation （DUSTM）, it can get better transmission performance in high transmission rate for MIMO-OFDM system.
文摘As the most influential trade platform for yarn and fiber industry players in Asia,Yarn Expo Autumn is renowned for its comprehensive range of exhibitors.As such,the show keeps gaining in popularity,with the number of exhibitors increasing from 258 to morethan 270 this year.To date,they come from 12 countries
文摘随着大语言模型(large language models,LLMs)(以下简称“大模型”)参数规模的持续增长,微调百亿级参数大模型对计算和存储资源提出了极高要求。传统分布式训练方案通常依赖大量高端GPU和高速互联网络,训练成本极为昂贵。现有单GPU训练方案虽通过张量卸载缓解显存压力,但仍然面临I/O传输效率低和设备利用率不足等问题。传统内核态I/O操作在大规模张量迁移中引入频繁的系统调用和上下文切换,成为制约性能的关键瓶颈;同时,优化器计算无法充分发挥多核CPU的并行能力,难以实现与GPU计算的有效重叠,进一步限制了系统性能。针对上述问题,提出了一种面向大模型训练的异构内存卸载与I/O优化方案HiTrain。首先构建了基于存储性能开发工具包(storage performance development kit,SPDK)的高性能张量存储模块,通过在用户态管理张量数据,避免了内核I/O栈开销,从而提高张量卸载的并发性与吞吐率;其次,设计并实现了基于异步优化器的存储-计算流水线调度模块,通过对优化器的执行进行优化重排来减少GPU等待时间,提高整体训练效率。实验结果表明,在配备单张GPU和非易失性存储器快速固态硬盘(non-volatile memory express solid state drive,NVMe SSD)的服务器上,所提出的方案能够充分利用系统中的存算资源,使得模型训练过程中张量卸载与加载效率提升32.7%,整体训练吞吐提升至现有方案的1.49倍,为低成本大模型训练提供了切实可行的技术路径。
文摘随机开关频率调制策略通过将开关频率在扩频范围内随机变化抑制高频谐波。虽然过宽的扩频范围能更有效地抑制高频谐波,但同时会带来严重的电流畸变和较大的转矩脉动。为此,该文提出在窄扩频范围下考虑随机数变化的随机开关频率空间矢量脉宽调制(random switching frequency space vector pulse width modulation,RSF-SVPWM)。与传统的RSF-SVPWM相比,提出的方法通过限制相邻随机数变化的最大值,减小了相邻随机数的相对变化,使高频谐波由三角形分布变为梯形分布,从而降低了高频谐波幅值,实现窄扩频范围内对高频谐波的有效抑制。实验结果表明,在相同的扩频范围内,该方法更有效地降低了开关频率及其整数倍处谐波的峰值。
