Based on the theory of language transfer in second language acquisition, this study explored the dynamic acquisition of English double object construction by Chinese EFL learners through comparative analysis. Altogeth...Based on the theory of language transfer in second language acquisition, this study explored the dynamic acquisition of English double object construction by Chinese EFL learners through comparative analysis. Altogether 120 subjects participated in this experiment and were required to take the proofreading exercise in limited time. The experimental results showed that Chinese EFL learners at different levels of proficiency acquired the core subclass of double object construction better than peripheral ones;meanwhile, learners at higher levels of proficiency outperformed those at lower levels, especially in the peripheral types. Relevant theoretical interpretations were given thereafter to the above research findings, with the hope to shed some light on the learning of double object construction by Chinese EFL learners.展开更多
井下运煤带式输送机是煤炭生产运输的关键环节。由于运煤皮带上出现的锚杆、槽钢、铁棍等异物,皮带在运行过程中容易出现纵向撕裂甚至断带等事故。针对煤矿井下的皮带异物检测问题,提出了一种基于迁移学习和在线难例挖掘的井下皮带异物...井下运煤带式输送机是煤炭生产运输的关键环节。由于运煤皮带上出现的锚杆、槽钢、铁棍等异物,皮带在运行过程中容易出现纵向撕裂甚至断带等事故。针对煤矿井下的皮带异物检测问题,提出了一种基于迁移学习和在线难例挖掘的井下皮带异物检测模型。首先,利用迁移学习策略,提高模型泛化能力,解决数据集较小的问题;其次,在特征融合层的改进型空间金字塔池化模块(Spatial Pyramid Pooling Fast,SPPF)添加坐标注意力(Coordinate Attention,CA)机制,提升模型特征的表达能力;之后,使用损失函数WIoU(Wise Intersection over Union)代替损失函数CIoU(Complete Intersection over Union),加快模型训练速度;最后,利用在线难例挖掘(Online Hard Example Mining,OHEM)策略,帮助模型更好地学习难分类的样本。试验结果表明,井下皮带异物检测模型在自建异物检测数据集上mAP@0.5和mAP@0.5-0.95分别取得了92.5%和79.4%的检测效果,与原YOLOv8相比分别增加了2.6百分点和1.8百分点,并且本模型在实际矿山的检测中取得了90.4%的检测效果,表明模型在实际矿井环境中具有较强的适用性,可为井下皮带异物的检测提供技术支持。展开更多
基于深度学习的目标检测算法已广泛应用,与此同时最近的一系列研究表明现有的目标检测算法容易受到对抗性攻击的威胁,造成检测器失效.然而,聚焦于自动驾驶场景下对抗攻击的迁移性研究较少,并且鲜有研究关注该场景下对抗攻击的隐蔽性.针...基于深度学习的目标检测算法已广泛应用,与此同时最近的一系列研究表明现有的目标检测算法容易受到对抗性攻击的威胁,造成检测器失效.然而,聚焦于自动驾驶场景下对抗攻击的迁移性研究较少,并且鲜有研究关注该场景下对抗攻击的隐蔽性.针对现有研究的不足,将对抗样本的优化类比于机器学习模型的训练过程,设计了提升攻击迁移性的算法模块.并且通过风格迁移的方式和神经渲染(neural rendering)技术,提出并实现了迁移隐蔽攻击(transferable and stealthy attack,TSA)方法.具体来说,首先将对抗样本进行重复拼接,结合掩膜生成最终纹理,并将其应用于整个车辆表面.为了模拟真实的环境条件,使用物理变换函数将渲染的伪装车辆嵌入逼真的场景中.最后,通过设计的损失函数优化对抗样本.仿真实验表明,TSA方法在攻击迁移能力上超过了现有方法,并在外观上具有一定的隐蔽性.此外,通过物理域实验进一步证明了TSA方法在现实世界中能够保持有效的攻击性能.展开更多
文摘Based on the theory of language transfer in second language acquisition, this study explored the dynamic acquisition of English double object construction by Chinese EFL learners through comparative analysis. Altogether 120 subjects participated in this experiment and were required to take the proofreading exercise in limited time. The experimental results showed that Chinese EFL learners at different levels of proficiency acquired the core subclass of double object construction better than peripheral ones;meanwhile, learners at higher levels of proficiency outperformed those at lower levels, especially in the peripheral types. Relevant theoretical interpretations were given thereafter to the above research findings, with the hope to shed some light on the learning of double object construction by Chinese EFL learners.
文摘井下运煤带式输送机是煤炭生产运输的关键环节。由于运煤皮带上出现的锚杆、槽钢、铁棍等异物,皮带在运行过程中容易出现纵向撕裂甚至断带等事故。针对煤矿井下的皮带异物检测问题,提出了一种基于迁移学习和在线难例挖掘的井下皮带异物检测模型。首先,利用迁移学习策略,提高模型泛化能力,解决数据集较小的问题;其次,在特征融合层的改进型空间金字塔池化模块(Spatial Pyramid Pooling Fast,SPPF)添加坐标注意力(Coordinate Attention,CA)机制,提升模型特征的表达能力;之后,使用损失函数WIoU(Wise Intersection over Union)代替损失函数CIoU(Complete Intersection over Union),加快模型训练速度;最后,利用在线难例挖掘(Online Hard Example Mining,OHEM)策略,帮助模型更好地学习难分类的样本。试验结果表明,井下皮带异物检测模型在自建异物检测数据集上mAP@0.5和mAP@0.5-0.95分别取得了92.5%和79.4%的检测效果,与原YOLOv8相比分别增加了2.6百分点和1.8百分点,并且本模型在实际矿山的检测中取得了90.4%的检测效果,表明模型在实际矿井环境中具有较强的适用性,可为井下皮带异物的检测提供技术支持。
文摘基于深度学习的目标检测算法已广泛应用,与此同时最近的一系列研究表明现有的目标检测算法容易受到对抗性攻击的威胁,造成检测器失效.然而,聚焦于自动驾驶场景下对抗攻击的迁移性研究较少,并且鲜有研究关注该场景下对抗攻击的隐蔽性.针对现有研究的不足,将对抗样本的优化类比于机器学习模型的训练过程,设计了提升攻击迁移性的算法模块.并且通过风格迁移的方式和神经渲染(neural rendering)技术,提出并实现了迁移隐蔽攻击(transferable and stealthy attack,TSA)方法.具体来说,首先将对抗样本进行重复拼接,结合掩膜生成最终纹理,并将其应用于整个车辆表面.为了模拟真实的环境条件,使用物理变换函数将渲染的伪装车辆嵌入逼真的场景中.最后,通过设计的损失函数优化对抗样本.仿真实验表明,TSA方法在攻击迁移能力上超过了现有方法,并在外观上具有一定的隐蔽性.此外,通过物理域实验进一步证明了TSA方法在现实世界中能够保持有效的攻击性能.
