基于雅克比稀疏自动编码机的手写数字识别算法

由于手写数字边缘轮廓差异大造成书写风格的不同,为了提高识别准确度,通过在自动编码机中加入稀疏约束项和雅克比正则项,提出一种雅克比稀疏自动编码机(JSAE)的算法进行手写数字识别。加入稀疏约束项能够有效提取数据中的隐藏结构,而雅克比正则化可以描述数据点的边缘特征,提高自动编码器算法的学习能力,从而更准确地抽取样本的本质特征。实验结果表明,JSAE在分类准确率上要高于自动编码机（AE）和稀疏自动编码机（SAE）算法。     

基于混合特征和堆栈稀疏自编码器的齿轮箱故障诊断

针对复杂工况下齿轮箱多故障信号诊断准确率低的问题,提出了一种基于混合特征和堆栈稀疏自编码器的齿轮箱故障诊断方法。从微观信号特征角度提取奇异值特征和小波分解后的样本熵特征;从宏观角度提取故障信号时域特征,将3种特征进行融合,并输入到由稀疏自编码和Softmax堆栈得到的深度神经网络中进行特征优化和分类识别。实验结果表明:在2种不同工况下,对6种齿轮箱故障数据进行诊断均表现出较高分类识别精度,且所构建的分类模型综合性能上均高于文中其他对比模型,因此本文作者所提出的方法能有效地进行齿轮箱故障诊断。     

基于数据降维与精确欧氏局部敏感哈希的k近邻推荐方法

针对基于k近邻的协同过滤推荐算法中存在的评分特征数据维度过高、k近邻查找速度慢,以及评分冷启动等问题,提出基于数据降维与精确欧氏局部敏感哈希（E²LSH）的k近邻协同过滤推荐算法。首先,融合评分数据、用户属性数据以及项目类别数据,将融合后的数据作为输入对堆叠降噪自编码（SDA）神经网络进行训练,取神经网络编码部分最后一个隐层的值作为输入数据的特征编码,完成非线性降维。然后,利用精确欧氏局部敏感哈希算法对降维后的数据建立索引,通过检索得到目标用户或目标项目的相似近邻。最后,计算目标与近邻之间的相似度,利用相似度对近邻的评分记录加权得到目标用户对目标项目的预测评分。在标准数据集上的实验结果表明,在冷启动场景下,均方根误差比基于局部敏感哈希的推荐算法（LSH-ICF）平均降低了约7.2%,平均运行时间和LSH-ICF相当。表明该方法在保证推荐效率的前提下,缓解了评分冷启动问题。     

基于非监督特征学习的兴趣点检测算法

由于兴趣点是图像中的基础、关键特征,因此兴趣点检测是图像配准、图像检索以及图像识别的关键步骤。基于兴趣点对于图像特征响应较为强烈的特性,结合非监督特征学习算法可以自主地从无标签的样本中提取特征的思想,提出了UFL-ID兴趣点检测算法。该算法无监督学习了图像的底层特征,对特征进行信息量和各向同性的评价,并利用特征的卷积响应及评价参数寻找图像中的兴趣点。与其他常见的兴趣点检测算法的对比实验表明,该算法具有良好的重复性与抗噪能力。     

基于监督学习深度自编码器的图像重构

针对数字图像受损信息的重构问题,提出一种将经典无监督学习自编码器(Auto-Encoder,AE)用于监督学习的新方法,并对深度模型结构与训练策略进行了研究。通过设计多组监督学习单层AE模型,提出了逐组“递进学习”和“关联编码”的学习策略,构建了一个新的基于监督学习的深度AE模型结构;对于新模型结构,采用多对一(一个输入样本的多种形式对应一个输出)的训练方法代替经典AE中一对一(一个输入样本对应一个输出)的训练方法。将该模型的结构和训练策略用于部分数据受损或遮挡的图像中进行数据重构测试,提高了模型对受损数据特征编码的表达能力和重构能力。实验结果表明,提出的新方法对于受损及遮挡样本的图像具有良好的重构效果和适应性。     

基于深度学习的边际Fisher分析特征提取算法

提取符合数据分布结构的特征一直是模式识别领域的热点问题。基于固定核映射方法具有获取非线性特征的能力,但对映射函数类型及其参数十分敏感。论文提出一种基于多层自动编码器的特征提取算法,该深度学习网络模型的训练分为无监督预训练以及基于边际Fisher准则的监督式精雕训练过程。通过数据生成性预训练和精雕过程中正则化手段防止过拟合训练。在多个数据集进行分类的实验结果进一步验证算法的有效性。     

新型稀疏自动编码器组合的深度学习方法

针对自动编码器在强噪声环境下分类效果低的特征,提出了基于改进型稀疏自动编码器组合的深度学习方法。在采用计算相关熵的方法,增强了稀疏自动编码器对非高斯噪声的鲁棒性的基础上,利用卷积神经网络对自动编码器进行边缘降噪,接着将改进后的稀疏自动编码器和边缘降噪自动编码器相结合,得到新的稀疏边缘降噪自动编码器。实测数据的实验结果表明,新的稀疏边缘降噪自动编码器比现有的分类算法,计算时间更短、准确率更高、效果更明显。     

An improved vector quantization method using deep neural network

To address the challenging problem of vector quantization (VQ) for high dimensional vector using large coding bits, this work proposes a novel deep neural network (DNN) based VQ method. This method uses a k-means based vector quantizer as an encoder and a DNN as a decoder. The decoder is initialized by the decoder network of deep auto-encoder, fed with the codes provided by the k-means based vector quantizer, and trained to minimize the coding error of VQ system. Experiments on speech spectrogram coding demonstrate that, compared with the k-means based method and a recently introduced DNN-based method, the proposed method significantly reduces the coding error. Furthermore, in the experiments of coding multi-frame speech spectrogram, the proposed method achieves about 11% relative gain over the k-means based method in terms of segmental signal to noise ratio (SegSNR).     

Automatic feature extraction of time-series applied to fault severity assessment of helical gearbox in stationary and non-stationary speed operation

Signals captured in rotating machines to obtain the status of their components can be considered as a source of massive information. In current methods based on artificial intelligence to fault severity assessment, features are first generated by advanced signal processing techniques. Then feature selection takes place, often requiring human expertise. This approach, besides time-consuming, is highly dependent on the machinery configuration as in general the results obtained for a mechanical system cannot be reused by other systems. Moreover, the information about time events is often lost along the process, preventing the discovery of faulty state patterns in machines operating under time-varying conditions. In this paper a novel method for automatic feature extraction and estimation of fault severity is proposed to overcome the drawbacks of classical techniques. The proposed method employs a Deep Convolutional Neural Network pre-trained by a Stacked Convolutional Autoencoder. The robustness and accuracy of this new method are validated using a dataset with different severity conditions on failure mode in a helical gearbox, working in both constant and variable speed of operation. The results show that the proposed unsupervised feature extraction method is effective for the estimation of fault severity in helical gearbox, and it has a consistently better performance in comparison with other reported feature extraction methods.     

Degradation prediction of proton exchange membrane fuel cell using auto-encoder based health indicator and long short-term memory network

As durability of proton exchange membrane fuel cell (PEMFC) remains as the main obstacle for its larger scale commercialization, predicting PEMFC degradation progress is thus an effective way to extend its lifetime. To realize reliable prediction, a novel health indicator (HI) extraction method based on auto-encoder is proposed in this paper, with which PEMFC future voltage can be predicted by long short-term memory network (LSTM). The effectiveness and robustness of proposed approach is investigated with test data simulating vehicle operation conditions, and accurate prediction performance can be observed, with the maximum root mean square error (RMSE) of 0.003513. Moreover, by comparing with two commonly prognostic methods including attention-based gated recurrent unit network and polarization model-LSTM, the proposed method can provide better predictions under various operating conditions. Furthermore, with the proposed method, the degradation mechanism of PEMFC can also be analyzed. Therefore, the proposed prognostic method can predict reliable PEMFC degradation progress and its corresponding degradation mechanisms, which will be beneficial in practical PEMFC systems for taking appropriate strategies to guarantee PEMFC durability.