当前,高职“Spark技术”课程教学存在学生前置知识不扎实、能力分化、考核方式缺乏针对性、项目案例孤立等问题,导致学生的技能与企业的需求脱节。文章结合项目驱动与成果导向教育(Outcome Based Education,OBE)理念,将生成式人工智能(A...当前,高职“Spark技术”课程教学存在学生前置知识不扎实、能力分化、考核方式缺乏针对性、项目案例孤立等问题,导致学生的技能与企业的需求脱节。文章结合项目驱动与成果导向教育(Outcome Based Education,OBE)理念,将生成式人工智能(Artificial Intelligence Generated Content,AIGC)融入课程教学,引入领域专用大语言模型SQLCoder-7B-2构建智能查询流程。教学案例借助Selenium技术获取租房信息,渗透“数据采集—数据清洗—数据分析”的大数据思维,详细阐述“课前—课中—课后”三阶段教学路径。该模式能有效降低Spark SQL编程学习难度,助力学生专注数据分析思维与AI工具应用能力培养,为大数据专业培育人工智能素养复合型人才提供可行教学范式。展开更多
Single-atom catalysts(SACs)are among the most cutting-edge catalysts in the multiphase catalysis track due to their unique geometrical and electronic properties,the highest atom utilization efficiency,and uniform acti...Single-atom catalysts(SACs)are among the most cutting-edge catalysts in the multiphase catalysis track due to their unique geometrical and electronic properties,the highest atom utilization efficiency,and uniform active sites.SACs have been facing an unresolved problem in practical applications:the opposing contradiction of activity-stability.The successful development of single-atom nano-islands(SANIs)cleverly combines the ultra-high atom utilization efficiency of SACs with the confinement effect and structural stability of nano-island structures,realizing the“moving but not aggregation”of SACs,which fundamentally solves this inherent contradiction.Although research on the precise loading of single atoms on nano-islands continues to advance,existing reviews have not yet established a closed-loop cognitive framework encompassing“models-synthesis-high stability mechanisms-high activity essence-applications.”This work fills this critical gap by systematically integrating the basic conceptual models and cutting-edge synthesis strategies of SANIs,focusing on revealing the underlying mechanisms by which SANIs overcome the stability bottleneck of SACs,elucidating the role of nano-islands and their synergistic mechanisms to clarify the high activity essence,and establishing the structure-activity relationship between atomic confinement effects and macroscopic performance,ultimately achieving breakthrough validation across catalytic systems.This review aims to open new perspectives,drive a paradigm shift in understanding the multi-dimensional advantages of SANIs,and thereby spur breakthrough progress in this frontier field.展开更多
This study investigated enhancing the wear resistance of Ti6Al4V alloys for medical applications by incorporating Ti C nanoreinforcements using advanced spark plasma sintering(SPS). The addition of up to 2.5wt% Ti C s...This study investigated enhancing the wear resistance of Ti6Al4V alloys for medical applications by incorporating Ti C nanoreinforcements using advanced spark plasma sintering(SPS). The addition of up to 2.5wt% Ti C significantly improved the mechanical properties, including a notable 18.2% increase in hardness(HV 332). Fretting wear tests against 316L stainless steel(SS316L) balls demonstrated a 20wt%–22wt% reduction in wear volume in the Ti6Al4V/Ti C composites compared with the monolithic alloy. Microstructural analysis revealed that Ti C reinforcement controlled the grain orientation and reduced the β-phase content, which contributed to enhanced mechanical properties. The monolithic alloy exhibited a Widmanstätten lamellar microstructure, while increasing the Ti C content modified the wear mechanisms from ploughing and adhesion(0–0.5wt%) to pitting and abrasion(1wt%–2.5wt%). At higher reinforcement levels, the formation of a robust oxide layer through tribo-oxide treatment effectively reduced the wear volume by minimizing the abrasive effects and plastic deformation. This study highlights the potential of SPS-mediated Ti C reinforcement as a transformative approach for improving the performance of Ti6Al4V alloys, paving the way for advanced medical applications.展开更多
Spark作为通用的计算引擎,以其简单、快速、可扩展的优势,被广泛地应用于大数据的处理和分析中.然而,Spark默认采用哈希分区或范围分区对数据进行划分,导致其在处理键倾斜分布的数据时,常常出现各分区数据量严重不均衡的问题.诸多优化...Spark作为通用的计算引擎,以其简单、快速、可扩展的优势,被广泛地应用于大数据的处理和分析中.然而,Spark默认采用哈希分区或范围分区对数据进行划分,导致其在处理键倾斜分布的数据时,常常出现各分区数据量严重不均衡的问题.诸多优化方法被提出,如迁移分区、贪心分区、反馈分区等,但往往存在数据传输量大、额外计算成本高、运行时间长等问题.为更好地缓解键倾斜分布问题带来的影响,本文提出了一种自适应的Spark数据均衡分区方法.该方法引入了奖惩思想对数据分区过程进行适当调控,同时对于数据量较大的键进行分割,使得各个分区的数据量相对均衡.该方法首先对数据采样并预估键权重.其次,按照键权重对样本数据降序排列,确保所有分区都有初始数据.再次,根据奖惩分配策略,自适应地更新各个分区的分配概率,并将待分配的键指向分配概率最高的分区.对于超过分区容量的键的数据,则分割为多个部分且指向不同分区.在所有样本数据分配完成后,获得自适应分区方案.在实际分区时,对于样本中出现的键对应的数据按照自适应分区方案进行分配;对于未出现的键对应的数据,则按照哈希方法进行分区.最后,通过实验验证,基于新方法设计的自适应均衡分区器(Adaptive Data Balanced Partitioner,ADBP)能够有效缓解键倾斜的负面影响.在真实数据集上,ADBP的WordCount程序总运行时间比自带分区器Hash、Range分别平均缩短了1.51%、29.90%,比现有基于学习自动机的自适应哈希分区器(Learning Automata Hash Partitioner,LAHP)、对倾斜的中间数据块进行拆分合并(Splitting and Combination algorithm for skew Intermediate Data block,SCID)算法、粗粒度放置和细粒度放置(Fined-Coarse Grained Intermediate Data Placement,FCGIDP)算法分别平均缩短了8.12%、21.64%、19.62%.展开更多
基金supported by the National Key Research and Development Program of China(No.2022YFD1700801-3)Key Laboratory of Jiangxi Province for Persistent Pollutants Prevention Control and Resource Reuse(No.2023SSY02061)support of the projects and research platform support provided by the laboratory.
文摘Single-atom catalysts(SACs)are among the most cutting-edge catalysts in the multiphase catalysis track due to their unique geometrical and electronic properties,the highest atom utilization efficiency,and uniform active sites.SACs have been facing an unresolved problem in practical applications:the opposing contradiction of activity-stability.The successful development of single-atom nano-islands(SANIs)cleverly combines the ultra-high atom utilization efficiency of SACs with the confinement effect and structural stability of nano-island structures,realizing the“moving but not aggregation”of SACs,which fundamentally solves this inherent contradiction.Although research on the precise loading of single atoms on nano-islands continues to advance,existing reviews have not yet established a closed-loop cognitive framework encompassing“models-synthesis-high stability mechanisms-high activity essence-applications.”This work fills this critical gap by systematically integrating the basic conceptual models and cutting-edge synthesis strategies of SANIs,focusing on revealing the underlying mechanisms by which SANIs overcome the stability bottleneck of SACs,elucidating the role of nano-islands and their synergistic mechanisms to clarify the high activity essence,and establishing the structure-activity relationship between atomic confinement effects and macroscopic performance,ultimately achieving breakthrough validation across catalytic systems.This review aims to open new perspectives,drive a paradigm shift in understanding the multi-dimensional advantages of SANIs,and thereby spur breakthrough progress in this frontier field.
文摘This study investigated enhancing the wear resistance of Ti6Al4V alloys for medical applications by incorporating Ti C nanoreinforcements using advanced spark plasma sintering(SPS). The addition of up to 2.5wt% Ti C significantly improved the mechanical properties, including a notable 18.2% increase in hardness(HV 332). Fretting wear tests against 316L stainless steel(SS316L) balls demonstrated a 20wt%–22wt% reduction in wear volume in the Ti6Al4V/Ti C composites compared with the monolithic alloy. Microstructural analysis revealed that Ti C reinforcement controlled the grain orientation and reduced the β-phase content, which contributed to enhanced mechanical properties. The monolithic alloy exhibited a Widmanstätten lamellar microstructure, while increasing the Ti C content modified the wear mechanisms from ploughing and adhesion(0–0.5wt%) to pitting and abrasion(1wt%–2.5wt%). At higher reinforcement levels, the formation of a robust oxide layer through tribo-oxide treatment effectively reduced the wear volume by minimizing the abrasive effects and plastic deformation. This study highlights the potential of SPS-mediated Ti C reinforcement as a transformative approach for improving the performance of Ti6Al4V alloys, paving the way for advanced medical applications.
文摘Spark作为通用的计算引擎,以其简单、快速、可扩展的优势,被广泛地应用于大数据的处理和分析中.然而,Spark默认采用哈希分区或范围分区对数据进行划分,导致其在处理键倾斜分布的数据时,常常出现各分区数据量严重不均衡的问题.诸多优化方法被提出,如迁移分区、贪心分区、反馈分区等,但往往存在数据传输量大、额外计算成本高、运行时间长等问题.为更好地缓解键倾斜分布问题带来的影响,本文提出了一种自适应的Spark数据均衡分区方法.该方法引入了奖惩思想对数据分区过程进行适当调控,同时对于数据量较大的键进行分割,使得各个分区的数据量相对均衡.该方法首先对数据采样并预估键权重.其次,按照键权重对样本数据降序排列,确保所有分区都有初始数据.再次,根据奖惩分配策略,自适应地更新各个分区的分配概率,并将待分配的键指向分配概率最高的分区.对于超过分区容量的键的数据,则分割为多个部分且指向不同分区.在所有样本数据分配完成后,获得自适应分区方案.在实际分区时,对于样本中出现的键对应的数据按照自适应分区方案进行分配;对于未出现的键对应的数据,则按照哈希方法进行分区.最后,通过实验验证,基于新方法设计的自适应均衡分区器(Adaptive Data Balanced Partitioner,ADBP)能够有效缓解键倾斜的负面影响.在真实数据集上,ADBP的WordCount程序总运行时间比自带分区器Hash、Range分别平均缩短了1.51%、29.90%,比现有基于学习自动机的自适应哈希分区器(Learning Automata Hash Partitioner,LAHP)、对倾斜的中间数据块进行拆分合并(Splitting and Combination algorithm for skew Intermediate Data block,SCID)算法、粗粒度放置和细粒度放置(Fined-Coarse Grained Intermediate Data Placement,FCGIDP)算法分别平均缩短了8.12%、21.64%、19.62%.
