ID 原文 译文
1813 针对上述问题，该文提出一种基于流量演化感知的服务功能链在线弹性编排策略(OEOP)， This paper proposes anonline elastic orchestration policy (OEOP)based on the evolution perception of flow rate to solve the above mentionedproblem.
1814 该策略将在线学习引入到 SFC 流量演化感知的过程，预先确定细粒度的 VNF 弹性需求。 OEOP introduces online learning into the evolution perception of flow rate, which helps to predetermine the fine-grained VNF scaling demands. In addition, the online elastic deployment is achieved according to the real-time update infor-mation of SFC paths and the load of nodes.
1815 此外，以实时更新的SFC 路径与节点负载两因子为导向，完成新增 VNF 的在线弹性部署，代替 VNF 迁移应对系统负载变化。 The newly deployed VNF instances can respond to the time-varying workload bytaking place of the mission of VNF migration.
1816 仿真表明，该策略明显增强了虚拟资源供应量与负载需求的匹配特性， The simulation results demonstrate that OEOP can significantly enhance the matching between virtual resource supply and workload demand.
1817 VNF 吞吐量利用率提高 10.2% ～ 24.8% ，运营开销平均降低 26.7% 。 The throughput of VNF is improved by 10. 2% ～ 24. 8% , and the operational expenditure can be reduced by 26. 7% on average compared with other solutions.
1818 双线性卷积网络(Bilinear CNN，B-CNN)在计算机视觉任务中有着广泛的应用。 The bilinear convolutional neural network(B-CNN)has been widely used in computer vision.
1819 B-CNN 通过对卷积层输出的特征进行外积操作，能够建模不同通道之间的线性相关，从而增强了卷积网络的表达能力。 B-CNN can capture the linear correlation between different channels by performing the outer product operation on the features of the con-volutional layer output, thus enhancing the representative ability of the convolutional network.
1820 由于没有考虑特征图中通道之间的非线性关系，该方法无法充分利用通道之间所蕴含的更丰富信息。 Since the non-linear relation-ship between the channels in the feature map is not taken account of, this method cannot make full use of the richer informa-tion contained between the channels.
1821 为了解决这一不足，本文提出了一种核化的双线性卷积网络，通过使用核函数的方式有效地建模特征图中通道之间的非线性关系，进一步增强卷积网络的表达能力。 In order to solve this problem, this paper proposes a kernelized bilinear convolutional neural network employing the kernel function to effectively capture the non-linear relationship between the channels in thefeature map, and further enhancing the representative ability of the convolutional network.
1822 本文在三个常用的细粒度数据库 CUB-200-2011、FGVC-Aircraft 以及 Cars 上对本文方法进行了验证，实验表明本文方法在三个数据库上均优于同类方法。 In this paper, the method is evalu-ated on three common fine-grained benchmarks CUB-200-2011, FGVC-Aircraft and Cars. Experiments show that our methodis superior to its counterparts on all three benchmarks.