ID 原文 译文
25765 针对该问题，本文提出了一种新的基于编码解码器结构的单幅图像去雨算法。 The new single image de-raining algorithm based on the coder and decoder structure is proposed in this paper．
25766 首先利用非局部操作获得不同像素点间的位置关系信息，从而获得图像全局信息表征。 Firstly, the positional rela-tionship information of various pixels between the points is obtained by non-local operation which can obtain the global in-formation of the image representation．
25767 其次，采用空间注意力机制对全局信息在空间维度位置上进行权值重标定，即在通道维度上对特征进行非线性建模，从而达到聚集相似特征和有用信息的目的。 Secondly, the spatial attention is applied to recalibrate the global information in the spatial dimension position, and the channel features are nonlinearly modeled in the channel dimension to aggregate similar characteristics and useful information．
25768 最后，利用反卷积与长距离残差连接逐层恢复去雨图像的大小。 Finally, the original size of rainy image is obtained by utilizing the de-convolution and long distance residual connection．
25769 分析和实验结果表明，本文提出算法雨痕去除效果明显，有效解决了去除具有不同雨密度大小雨条纹的现实困难，同时较好地保留图像的细节和边缘信息。 The experimental results and analysis show that the proposed algorithm can obtain significant de-raindrop effect. The proposed algorithm can also resolve the difficulties of removing raindrops with various densities while maintaining the details of the image and edge information．
25770 机器学习的飞速发展使其成为数据挖掘领域最有效的工具之一，但算法的训练过程往往需要大量的用户数据，给用户带来了极大的隐私泄漏风险。 The rapid development of machine learning makes itself one of the most effective tools in the data mining research community. However, the training of algorithm often needs a large amount of user data, which brings a great risk of privacy leakage to users．
25771 由于数据统计特征的复杂性及语义丰富性，传统隐私数据发布方法往往需要对原始数据进行过度清洗，导致数据可用性低而难以再适用于数据挖掘任务。 Due to the complex statistical characteristics and semantic richness of the data, traditional privatedata publishing methods usually sanitize original data too excessively to lead to low data availability and uselessness in datamining tasks．
25772 为此，提出了一种基于生成对抗网络( Generative Adversarial Network，GAN) 的差分隐私数据发布方法，通过在 GAN 模型训练的梯度上添加精心设计的噪声来实现差分隐私，确保 GAN 可无限量生成符合源数据统计特性且不泄露隐私的合成数据。 In this paper, a differential private data publishing method based on generative adversarial network ( GAN) is proposed. The differential privacy of the GAN model is realized by adding carefully designed noise to the gradients during the training procedure, so that the GAN can generate unlimited synthetic data conforming to the original statistical characteristics without disclosing any privacy．
25773 针对现有同类方法合成数据质量低、模型收敛缓慢等问题，设计多种优化策略来灵活调整隐私预算分配并减小总体噪声规模，同时从理论上证明了合成数据严格满足差分隐私特性。 Aiming at the problems of low quality synthetic data and slow convergence in the existing similar methods, several optimization strategies are designed to adjust the privacy budget allocation and reduce the overall noise scale． Moreover, we provide rigorous proof that the synthetic data satisfies the differential privacy.
25774 在公开数据集上与现有方法进行实验对比，结果表明本方法能够更高效地生成质量更高的隐私保护数据，适用于多种数据分析任务。 Comparisons with existing methods on public datasets show that the method proposed can generate private data with higher quality more efficiently, which is suitable for various data analysis tasks．