ID	原文	译文
3533	为了改善电离层频率预报准确性低、短波通信连通率差以及工程中观测野值对预报精度影响严重等问题，提出一种基于稳健卡尔曼滤波的倾斜探测电离层 MUF 短期预报方法。	To improve the low accuracy of ionospheric frequency forecasting, poor HF communication connectivity, and the severe impact of observational outliers in engineering on forecast accuracy, a short-term prediction method foroblique detection ionospheric MUF based on robust Kalman filtering was proposed.
3534	通过研究实测 MUF 数据的变化规律及观测误差的成因与特征，利用电离层参考模型作为先验信息，改进卡尔曼滤波模型的状态估计方程；	By studying the variation law of themeasured MUF data and the causes as well as characteristics of observation errors, and exploiting the ionospheric refer-ence model as a priori information, the state estimation equation of the Kalman filter model was improved.
3535	引入代价函数机制，通过 Huber-M 估计实现对预报状态的量测更新，减少观测野值对预报结果的影响，提高所提方法的稳健性。	The cost func-tion mechanism was introduced to realize the measurement update of the forecast state through Huber-M estimation,which reduced the influence of the observation outliers on the forecast results and improves the robustness of the pro-posed method.
3536	仿真结果表明，所提方法能有效抑制观测野值带来的不利影响，具有较强的稳健性和稳定性。	The simulation results show that the proposed method can effectively suppress the adverse effects causedby the observation outliers and has good robustness and stability.
3537	为了解决已有 YOLOv3 算法对于存在小目标问题和背景复杂问题的交通标志检测任务会有较多的误检和漏检的问题，在 YOLOv3 算法的基础上，提出了目标检测的通道注意力方法和基于语义分割引导的空间注意力方法，形成 YOLOv3-A 算法。	To solve the problem that the existing YOLOv3 algorithm had more false detections and missed detections fortraffic sign detection task with small target problems and complex background, based on the YOLOv3, a channel atten-tion method for target detection and a spatial attention method based on semantic segmentation guidance were proposedto form the YOLOv3-A (attention) algorithm.
3538	YOLOv3-A 算法通过对检测分支特征在通道和空间 2 个维度进行重新标定，使网络聚焦和增强有效特征，并抑制干扰特征，提高了算法的检测能力。	The detection features in the channel and spatial dimensions were recali-brated, allowing the network to focus and enhance the effective features, and suppress interference features, which greatly improved the detection performance.
3539	在 TT100K 交通标志数据集上的实验表明，所提算法对小目标检测性能的改善尤为明显，相比于 YOLOv3 算法，所提算法的精度和召回率分别提升了 1.9%和 2.8%。	Experiments on the TT100K traffic sign data set show that the algorithm improvesthe detection performance of small targets, and the accuracy and recall rate of the YOLOv3 are improved by 1.9% and2.8% respectively.
3540	针对汉语谓语中心词识别困难及唯一性的问题，提出了一种基于 Highway-BiLSTM 网络的深度学习模型。	Aiming at the problem of difficult recognition and uniqueness of Chinese predicate head, a Highway-BiLSTMmodel was proposed.
3541	首先，通过多层 BiLSTM 网络叠加获取句子内部不同粒度抽象语义信息的直接依赖关系；	Firstly, multi-layer BiLSTM networks were used to capture multi-granular semantic dependence in asentence.
3542	然后，利用 Highway网络缓解深层模型出现的梯度消失问题；	Then, a Highway network was adopted to alleviate the problem of gradient disappearance.