We consider in this paper the boundary value problems of nonlinear systems the form εY″=F(t,Y,Y′,ε), -1<t<1, Y(-1,ε)=A(ε), Y(1,ε)=B(ε). Supoosing some or all of the components of F , that is, ...We consider in this paper the boundary value problems of nonlinear systems the form εY″=F(t,Y,Y′,ε), -1<t<1, Y(-1,ε)=A(ε), Y(1,ε)=B(ε). Supoosing some or all of the components of F , that is, f i satisfy 2 f y′ 2 i t =0 =0, we say that F possesses a generalized turning point at t =0. Our goal is to give sufficient conditions for the existence of solution of the problems and to study the asymptotic behavior of the solution when F possesses a generalized turning point at t =0. We mainly discuss regular singular crossings.展开更多
针对绝缘子缺陷检测算法具有较大的参数规模和计算量导致难以部署在边缘设备,模型剪枝后难以获得正确连接,且过度稀疏化训练导致模型精度大幅度下降等问题,提出一种基于DepGraph偏移正则化的绝缘子多缺陷检测轻量化算法。通过依赖图(Dep...针对绝缘子缺陷检测算法具有较大的参数规模和计算量导致难以部署在边缘设备,模型剪枝后难以获得正确连接,且过度稀疏化训练导致模型精度大幅度下降等问题,提出一种基于DepGraph偏移正则化的绝缘子多缺陷检测轻量化算法。通过依赖图(DepGraph)对改进后YOLOv7网络建立连接关系模型,再添加偏移正则化稀疏约束对其进行组级的稀疏训练,删除冗余的连接,得到参数规模和计算量更小的轻量型检测算法。将提出的模型压缩算法应用到绝缘子多缺陷检测任务中,实验结果表明,剪枝后模型相较于未剪枝模型的参数规模和计算量分别下降65.25%和65.98%,而平均准确率仅减少1.1个百分点,验证了DepGraph偏移正则化方案在绝缘子多缺陷检测任务中的有效性;在CIFAR-10数据集上进行实验,实验结果表明,在加速比为2.88时,所提算法仍可以保持93.69%的分类精度。使用TensorRT对该算法进行推理加速,并在Jetson Orin Nano平台上部署,经过TensorRT优化后模型的检测速度达到了35.24帧/s,符合在移动设备上部署的需求。展开更多
文摘We consider in this paper the boundary value problems of nonlinear systems the form εY″=F(t,Y,Y′,ε), -1<t<1, Y(-1,ε)=A(ε), Y(1,ε)=B(ε). Supoosing some or all of the components of F , that is, f i satisfy 2 f y′ 2 i t =0 =0, we say that F possesses a generalized turning point at t =0. Our goal is to give sufficient conditions for the existence of solution of the problems and to study the asymptotic behavior of the solution when F possesses a generalized turning point at t =0. We mainly discuss regular singular crossings.
文摘针对绝缘子缺陷检测算法具有较大的参数规模和计算量导致难以部署在边缘设备,模型剪枝后难以获得正确连接,且过度稀疏化训练导致模型精度大幅度下降等问题,提出一种基于DepGraph偏移正则化的绝缘子多缺陷检测轻量化算法。通过依赖图(DepGraph)对改进后YOLOv7网络建立连接关系模型,再添加偏移正则化稀疏约束对其进行组级的稀疏训练,删除冗余的连接,得到参数规模和计算量更小的轻量型检测算法。将提出的模型压缩算法应用到绝缘子多缺陷检测任务中,实验结果表明,剪枝后模型相较于未剪枝模型的参数规模和计算量分别下降65.25%和65.98%,而平均准确率仅减少1.1个百分点,验证了DepGraph偏移正则化方案在绝缘子多缺陷检测任务中的有效性;在CIFAR-10数据集上进行实验,实验结果表明,在加速比为2.88时,所提算法仍可以保持93.69%的分类精度。使用TensorRT对该算法进行推理加速,并在Jetson Orin Nano平台上部署,经过TensorRT优化后模型的检测速度达到了35.24帧/s,符合在移动设备上部署的需求。
