ID 原文 译文
25025 该模型在网络的残差映射结构中加入残差混合注意力模块，解决了原模型提取的特征分辨性弱的问题，精确捕捉了脑部组织在 CT 图像中的位置和内容信息; A residual hybrid attention module is embedded in the residual identity mapping to capture the location and content information of brain tissue in brain CT images, solving the original model to extract weak distinguish features problems．
25026 此外，本文设计了全局平均池化层，简化了模型的复杂度，并在其后引入 Dropout 机制，缓解了过拟合。 In addition, to simplify the improved model and alleviate the overfitting, several techniques such as global average pooling and Dropout are used in the model．
25027 在训练阶段，该模型建立了标签平滑交叉熵损失函数，使模型在样本数量有限的情况下仍有较强的泛化能力。 Moreover, to have strong generalization ability in the case of limited sample quantity, tag smoothing cross-entropy loss function is adopted to train the model．
25028 系列实验证明了改进后的 ＲesNet-10 网络模型在分类脑部 CT 图像时达到 97. 47% 的分类精度。 Experimental results show that the improved ＲesNet-10 achieves 97.47% accuracy in classifying brain CT images．
25029 本文针对灰狼优化(Grey Wolf Optimizer，GWO)算法平衡全局探索和局部搜索能力的不足，提出了一种基于反向改进的灰狼算法(Opposition Learning Grey Wolf Optimizer，OLGWO)，来优化预测模型的超参数，以提高其用于交通流预测的精度与鲁棒性。 Focusing on the weakness of the balance capability between global exploration and local search in the classical GWO algorithm, a novel grey wolf algorithm based on opposition learning (OLGWO), which can evolve the hyper-parameters of forecasting model, is proposed to improve the accuracy and enhance the robustness of traffic flow forecasting models.
25030 本算法在迭代过程中采用了反向学习策略，并引入了等级相关概念，主要通过计算普通狼与目标狼的 Spearman 相关系数，并根据其值来选择性地更新狼种群。 This algorithm is designed to take advantage of opposition learning strategy with the iterative process, and exploits the concept of rank correlation that can describe the Spearman correlation coefficients between the target wolf and the common wolves, and then selectively updates the each wolf of the whole population according to their values．
25031 实验先对 12 个标准测试函数对比了四种算法 OLGWO、TGWO(Transformed Grey Wolf Optimizer)、GWO、PSO(Particle Swarm Optimization)，得到了寻优均值和标准差，验证了 OLGWO 算法具有突出的性能优势; Firstly, the performance comparison of four algorithms (OLGWO, TGWO, GWO, PSO), based on 12 benchmark functions, is conducted in terms of the two metrics, namely the optimization means and standard deviations. The results verify the outstanding performance of the proposed algorithm．
25032 然后采用美国加州公路交通流数据，在不同缺失率下比较了四种算法优化的反向传播(Back Propagation，BP)网络模型，结果显示，OLGWO-BP 模型预测精度比其它三种模型最高分别有1.95% 、3.98% 和 11.07% 的提升，同时表现出更好的稳定性。 Furthermore, based on the California highway traffic flow data, the four models optimized by the concerned algorithms are compared under different loss rates. The results show that the prediction accuracy of OLGWO-BP is higher than that of the others by 1.95% , 3.98% and 11.07% , respectively, and the stability is better．
25033 针对现有可信中继 QKD(Quantum Key Distribution)网络路由方案应用于广域环境时存在着密钥交换效率低、密钥资源无意义消耗大的问题，分析了影响密钥交换效率的因素，设计了适应广域 QKD 网络的分层路由方案。 Aiming at the problems, such as low key exchange efficiency and large meaningless consumption of secret key materials, when the existing routing schemes for trust relaying QKD (quantum key distribution) network used in the wide-area environment, a hierarchical routing scheme which is suitable for wide-area QKD network was designed.
25034 该方案将 QKD 网络划分为若干路由域，并通过拓扑聚合构建分层 QKD 网络，设计了基于最低层网络匹配的跨域密钥路由算法，使得高层路由域内一跳便可跨过多个低层路由域，极大地降低了密钥中继跳数，提高了密钥交换效率及密钥资源利用率。 This routing scheme divided the QKD network into multiple routing areas, built a hierarchical network by topological aggregation and designed a cross-domain routing algorithm based on the principle of the lowest layer matching． Then the hop number in the routing path is decreased, and the efficiency of key exchange and the utilization rate of secret key materials ware increased.