共查询到10条相似文献，搜索用时 46 毫秒
1.
Error detection is a critical step in data cleaning. Most traditional error detection methods are based on rules and external information with high cost, especially when dealing with large-scaled data. Recently, with the advances of deep learning, some researchers focus their attention on learning the semantic distribution of data for error detection; however, the low error rate in real datasets makes it hard to collect negative samples for training supervised deep learning models. Most of the existing deep-learning-based error detection algorithms solve the class imbalance problem by data augmentation. Due to the inadequate sampling of negative samples, the features learned by those methods may be biased. In this paper, we propose an AEGAN (Auto-Encoder Generative Adversarial Network)-based deep learning model named SAT-GAN (Self-Attention Generative Adversarial Network) to detect errors in relational datasets. Combining the self-attention mechanism with the pre-trained language model, our model can capture semantic features of the dataset, specifically the functional dependency between attributes, so that no rules or constraints are needed for SAT-GAN to identify inconsistent data. For the lack of negative samples, we propose to train our model via zero-shot learning. As a clean-data tailored model, SAT-GAN tries to recognize error data as outliers by learning the latent features of clean data. In our evaluation, SAT-GAN achieves an average -score of 0.95 on five datasets, which yields at least 46.2% -score improvement over rule-based methods and outperforms state-of-the-art deep learning approaches in the absence of rules and negative samples. 相似文献
2.
为了提高断层识别的准确率,提出改进Unet模型.为编码器部分设计一种多分支的并联结构M-block(Multi-branch block),它可以捕获多尺度上下文信息,并且多分支的并联结构会带来高性能收益.在解码器部分加入Self-Attention块和注意力门控机制.Self-Attention通过对输入特征上下文的加权平均操作,不仅使注意力模块能够灵活地关注图像的不同区域,而且弥补了 CNN(Convolutional Neural Network)网络局部性的缺点,为神经网络带来更多的可能性.通过合成数据和实际数据证实,该模型将传统卷积中的权值共享优点和Self-Attention动态计算注意力权重的优点结合,提高了断层识别的精度,与Unet相比,验证损失下降了 33.68％.模型不仅准确识别出了断层特征,且比目前流行的深度学习方法更准确. 相似文献
3.
4.
针对实际鸟类监测环境中,收集鸟声声频数据分布不均匀,导致神经网络训练不充分,分类识别测试准确率低的问题,设计了一种桥接Transformer神经网络模型。该网络首先利用原始鸟声声频信号生成短时傅里叶变换语谱图作为输入特征,之后将语谱图输入到由注意力模块和卷积模块桥接组成的Transformer网络中,完成对语谱图中全局特征和局部特征的信息交互,最后利用单层Transformer编码器实现对每一个批次样本的损失优化,得到最终的分类结果。在Birdsdata和xeno-canto鸟声数据集上进行小样本实验,分别获得了91.34%和82.63%的平均准确率,与其他鸟声识别网络进行了对比实验,验证了该网络的有效性。 相似文献
5.
In this paper we investigate the optical properties of an open four-level tripod atomic system driven by an elliptically polarized probe field in the presence of the external magnetic field and compare its properties with the corresponding closed system.Our result reveals that absorption,dispersion and group velocity of probe field can be manipulated by adjusting the phase difference between the two circularly polarized components of a single coherent field,magnetic field and cavity parameters i.e.the atomic exit rate from cavity and atomic injection rates.We show that the system can exhibit multiple electromagnetically induced transparency windows in the presence of the external magnetic field.The numerical result shows that the probe field in the open system can be amplified by appropriate choice of cavity parameters,while in the closed system with introduce appropriate phase difference between fields the probe field can be enhanced.Also it is shown that the group velocity of light pulse can be controlled by external magnetic field,relative phase of applied fields and cavity parameters.By changing the parameters the group velocity of light pulse changes from subluminal to superluminal light propagation and vice versa. 相似文献
6.
1引言换热器是应用于多种工业系统中的重要设备。随着节能研究的深入，对各种换热器的性能要求愈来愈高，因此对换热器的设计方法也提出了更高要求。过增元等［‘l提出了“换热器温差场愈均匀，不仅热力学效益愈高，而且传热效能也愈高”的新构想，并称其为“温差场均匀性原则”。文献山用分析和数值计算的方法进行了初步的证明。据我们掌握的资料，尚未有对温差场均匀性原则直接证明的文献发表。凝汽器是火电厂最重要的换热设备。目前大型汽轮机组已广泛采用双背压或多背压凝汽器以提高蒸汽动力循环的热效率、杨善让等［’l在假定各级热负… 相似文献
7.
Parameter identification of chaos system based on unknown parameter observer is discussed generally. Based on the work of Guan et al. [X.P. Guan, H.P. Peng, L.X. Li, et al., Acta Phys. Sinica 50 (2001) 26], the design of unknown parameter observer is improved. The application of the improved approach is extended greatly. The works in some literatures [X.P. Guan, H.P. Peng, L.X. Li, et al., Acta Phys. Sinica 50 (2001) 26; J.H. Lü, S.C. Zhang, Phys. Lett. A 286 (2001) 148; X.Q. Wu, J.A. Lu, Chaos Solitons Fractals 18 (2003) 721; J. Liu, S.H. Chen, J. Xie, Chaos Solitons Fractals 19 (2004) 533] are only the special cases of our Corollaries 1 and 2. Some observers for Lü system and a new chaos system are designed to test our improved method, and simulations results demonstrate the effectiveness and feasibility of the improved approach. 相似文献
8.
八种鹅膏菌的傅里叶变换红外光谱的差谱鉴别研究 总被引：1，自引：0，他引：1
利用傅里叶变换红外光谱技术对长柄鹅膏菌、粗鳞白鹅膏菌、格纹鹅膏菌、红黄鹅膏菌、黄柄鹅膏菌、灰疣鹅膏菌、欧氏鹅膏菌、小豹斑鹅膏菌进行傅里叶变换红外光谱研究, 发现八种鹅膏菌的傅里叶变换红外光谱极为相似, 特征区和指纹区(1800~1100 cm-1)的相关系数均大于0.966。因此从这八种鹅膏菌的原始光谱对其鉴别将十分困难。通过差谱技术处理后, 八种鹅膏菌在1800~1100 cm-1范围呈现出各自的特征, 相关分析结果定量反映出它们之间差异较为明显。利用差谱中特征区和指纹区的差异可快速鉴别出该八种鹅膏菌。研究表明: 傅里叶变换红外光谱技术能提供大型真菌所含化学成分的分子结构信息, 结合差谱技术可以鉴别同属下的不同种高等真菌。 相似文献
9.
