This paper studies synchronization of all nodes in a fractional-order complex dynamic network. An adaptive control strategy for synchronizing a dynamic network is proposed. Based on the Lyapunov stability theory, this...This paper studies synchronization of all nodes in a fractional-order complex dynamic network. An adaptive control strategy for synchronizing a dynamic network is proposed. Based on the Lyapunov stability theory, this paper shows that tracking errors of all nodes in a fractional-order complex network converge to zero. This simple yet prac- tical scheme can be used in many networks such as small-world networks and scale-free networks. Unlike the existing methods which assume the coupling configuration among the nodes of the network with diffusivity, symmetry, balance, or irreducibility, in this case, these assumptions are unnecessary, and the proposed adaptive strategy is more feasible. Two examples are presented to illustrate effectiveness of the proposed method.展开更多
智能工单系统是企业数字化转型的核心支撑平台。当前,智能工单系统面临多源异构数据的跨模态冗余及语义冲突问题,传统基于单模态处理架构的大语言模型(Large Language Model,LLM)存在关键信息漏检率高、数据清洗效果差的缺陷,严重制约...智能工单系统是企业数字化转型的核心支撑平台。当前,智能工单系统面临多源异构数据的跨模态冗余及语义冲突问题,传统基于单模态处理架构的大语言模型(Large Language Model,LLM)存在关键信息漏检率高、数据清洗效果差的缺陷,严重制约了工单系统智能化发展。针对该问题,混合专家模型(Mixture of Experts,MoE)可通过动态路由机制自适应分配多模态数据至特定专家网络,在提升跨模态特征融合精度的同时显著优化计算效率。基于此,提出一种基于混合专家模型的多模态工单数据智能处理方法。首先基于DeepSeekMoE架构设计了一种语义分析模型,以实现跨模态数据的特征解耦与关键内容提取。其次提出基于Thinker-Talker的多模态特征融合架构,有效提升冗余数据利用率与语义一致性。最后设计非结构化数据清洗与结构化表单生成算法,完成原始数据的降噪清洗与语义增强,输出符合规范的结构化工单。消融实验表明,该方法在私有化数据集上的信息提取精度达92.7%,较传统工单处理方式的标准符合度提升36.2%,为智能工单系统多模态数据处理提供了可扩展的技术范式。展开更多
基金Project supported by the National Natural Science Foundation of China(Nos.11672231 and11672233)the Natural Science Foundation of Shaanxi Province(No.2016JM1010)+1 种基金the Fundamental Research Funds for the Central Universities(No.3102017AX008)the Seed Foundation of Innovation and Creation for Graduate Students at the Northwestern Polytechnical University of China(No.Z2017187)
文摘This paper studies synchronization of all nodes in a fractional-order complex dynamic network. An adaptive control strategy for synchronizing a dynamic network is proposed. Based on the Lyapunov stability theory, this paper shows that tracking errors of all nodes in a fractional-order complex network converge to zero. This simple yet prac- tical scheme can be used in many networks such as small-world networks and scale-free networks. Unlike the existing methods which assume the coupling configuration among the nodes of the network with diffusivity, symmetry, balance, or irreducibility, in this case, these assumptions are unnecessary, and the proposed adaptive strategy is more feasible. Two examples are presented to illustrate effectiveness of the proposed method.
文摘智能工单系统是企业数字化转型的核心支撑平台。当前,智能工单系统面临多源异构数据的跨模态冗余及语义冲突问题,传统基于单模态处理架构的大语言模型(Large Language Model,LLM)存在关键信息漏检率高、数据清洗效果差的缺陷,严重制约了工单系统智能化发展。针对该问题,混合专家模型(Mixture of Experts,MoE)可通过动态路由机制自适应分配多模态数据至特定专家网络,在提升跨模态特征融合精度的同时显著优化计算效率。基于此,提出一种基于混合专家模型的多模态工单数据智能处理方法。首先基于DeepSeekMoE架构设计了一种语义分析模型,以实现跨模态数据的特征解耦与关键内容提取。其次提出基于Thinker-Talker的多模态特征融合架构,有效提升冗余数据利用率与语义一致性。最后设计非结构化数据清洗与结构化表单生成算法,完成原始数据的降噪清洗与语义增强,输出符合规范的结构化工单。消融实验表明,该方法在私有化数据集上的信息提取精度达92.7%,较传统工单处理方式的标准符合度提升36.2%,为智能工单系统多模态数据处理提供了可扩展的技术范式。
