Results 21 to 30 of about 1,128,392 (212)
Zhejiang Daxue xuebao. Lixue ban, 2023 提出了一种新颖的广义无监督函数映射学习的三维形状密集对应方法。首先，基于多层感知器（multilayer perceptron，MLP）和残差网络，直接学习深度点特征。其次，计算点云的近似测地线距离，并对其进行特征分解，建立特征嵌入空间，引入注意力机制，有效学习广义基函数表示。再次，结合点特征与广义基函数生成三维形状的深度特征表示。最后，建立无监督的函数映射网络框架，获取形状之间的密集对应表示。提出的三元正则优化机制，联合重构损失、特征损失和形状匹配的距离损失 ...
窦丰(DOU Feng) +6 more
doaj +1 more source
, 2004 为了考察注意能力的年龄差异,分别从注意分配能力和注意选择能力两个方面对青年和老年被试的表现进行了比较.结果表明:青、老年在注意分配能力上没有显著的差异;而青、老年在注意选择能力上却有显著的差异,老年的注意选择能力有显著的衰退.因此 ...
焦书兰, 罗婷
core +2 more sources
基于双重注意力变换模型的分布式屋顶光伏变电站级日前功率预测 [PDF]
全球能源互联网分布式屋顶光伏地理位置分散，受地理环境遮挡和多种气象因素影响，导致光伏出力特性存在差异，给变电站级分布式屋顶光伏日前功率预测造成挑战。针对上述问题，提出了一种基于双重注意力变换模型的分布式屋顶光伏变电站级日前功率预测方法。首先，基于动态时间规整算法计算分布式光伏用户出力特性间的相似度，并基于凝聚层次聚类法将其划分成若干类；然后，利用自主注意力网络学习各时间步间的时序关联特性，通道卷积注意力机制学习多特征变量间的相关性，构建日前功率预测模型；最后，将每一类日前预测结果相加，实现变电站级日前功率预测 ...
王光华 +5 more
doaj +1 more source
International Journal of Applied Linguistics, EarlyView.ABSTRACT Multilingual students in Anglophone universities often operate in survival mode. While translanguaging supports learning, critical gaps remain in understanding how translanguaging pedagogies transform and sustain motivation in English‐dominant contexts.
Melissa Jufenna Slamet, Julie Choi
wiley +1 more source
International Journal of Applied Linguistics, EarlyView.ABSTRACT This study examines how foreign language teachers’ beliefs about the medium of instruction and translanguaging evolve over time by revisiting three in‐service teachers twelve years after their initial interviews. Using a qualitative longitudinal design, the study traces how translanguaging was initially framed as pedagogically deficient but ...
Danping Wang
wiley +1 more source
Dizhen xuebao为了提高频散曲线反演算法的精度，本文提出一种将自注意力机制与宽度学习相结合的瑞雷面波频散曲线反演方法。该方法在宽度学习系统的基础上引入自注意力机制，通过注意力加权的方式对提取到的面波特征进行融合与变换，以增强宽度学习系统的特征学习能力，进而实现面波频散曲线的更高精度反演。然后对典型地层模型生成的理论合成数据进行反演测试，结果显示本文所提智能反演算法可高效完成面波频散曲线反演。最后通过实际资料的反演测试验证该方法能够获得地层横波速度的高精度估计结果，有望在工程勘察与浅层地质勘探中得到广泛应用。
Xianna Lu +6 more
doaj +1 more source
International Journal of Applied Linguistics, EarlyView.ABSTRACT While English‐medium instruction (EMI) and English for Academic Purposes (EAP) courses are designed to promote language and content learning, their prevailing English‐only approaches often undermine students’ learning motivation. Against this backdrop, translanguaging has gained recognition as an effective pedagogical practice to support ...
Yihai Wen, Fan Fang
wiley +1 more source
Gaoya dianqi, 2022 绝缘子串是输电线路系统中至关重要的器件之一，基于图像处理的绝缘子串识别，能够更快更准确地识别到绝缘子串。针对输电线路中绝缘子串与绝缘子串、绝缘子串与背景之间相互干扰导致目标识别时存在误差较大的问题，提出一种融入注意力机制的YOLOv3绝缘子串识别方法。该方法在YOLOv3模型的基础上，首先在特征提取网络Darknet-53中加入无参注意力SimAM，聚焦网络和增强有效特征，抑制干扰特征，提高网络对绝缘子串的注意能力；其次根据绝缘子串的形状特征，调整模型预设Anchor box的值；最后以焦点损失 ...
李季 +4 more
doaj
Scientists in power plays: How substantive were scientists' narratives during the COVID‐19 pandemic?
Policy Studies Journal, Volume 54, Issue 3, August 2026.Abstract Scientists who provide expert advice must engage with policy processes. Little is known about whether and how scientists deal with the political dynamics of policy processes that are inherent to policymaking. We study this question by building on the policy dimension concept within the Narrative Policy Framework.
Jule Ksinsik, Caroline Schlaufer
wiley +1 more source
基于改进CNN-BiLSTM模型和地磁监测数据的多时间长度GIC预测
Kongjian kexue xuebao太阳风暴在电力系统网络中驱动产生的GIC会影响电力设备和系统的安全运行, 严重时还会引发大面积停电事件. 预测电网GIC水平能够为电力系统保护措施提供重要参考, 然而对这方面的研究仍显不足. 为了解决该问题, 将卷积神经网络(CNN)与双向长短时记忆(BiLSTM)以及注意力机制相结合, 利用空间天气的相关监测信息, 提出了大规模电网GIC多时间长度的预测方法.
蓝 东亮 +6 more
doaj +1 more source