Search

Copyright of Journal of Southeast University / Dongnan Daxue Xuebao is the property of Journal of Southeast University Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In the realm of natural language processing (NLP), the use of pre-trained models has seen a significant rise in practical applications. These models are initially trained on extensive datasets, encompassing both monolingual and multilingual data, and can be subsequently fine-tuned for target output using a smaller, task-specific dataset. Recent research in multilingual neural machine translation (NMT) has shown potential in creating architectures that can incorporate multiple languages. One such model is mBART50, which was trained on 50 different languages. This paper presents a work on fine-tuning mBART50 for NMT in the absence of high-quality bitext. Adapting a pre-trained multilingual model can be an effective approach to overcome this challenge, but it may not work well when the translation pairs contain languages not seen by the pre-trained model. In this paper, the resilience of the self-supervised multilingual sequence-to-sequence pre-trained model (mBART50) were investigated when fine-tuned with small amounts of high-quality bitext or large amounts of noisy parallel data (Kazakh-Russian). It also shows how mBART improves a neural machine translation system on a low-resource translation pair, where at least one language is unseen by the pre-trained model (Russian-Tatar). The architecture of mBART was employed in this study, adhering to the traditional sequence-to-sequence Transformer design. A Transformer Encoder-Decoder model with Byte Pair Encoding (BPE) was trained in our baseline experiment. The experiments show that fine-tuned mBART models outperform Baseline Transformer-based NMT models in all tested translation pairs, including cases where one language is unseen during mBART pretraining. The results show an increase in the BLEU score of 11.95 when translating from Kazakh to Russian and by 1.17 points in BLEU score when translating from Russian to Tatar. Utilizing pre-trained models like mBART can substantially reduce the data and computational requirements for NMT, leading to improved translation performace for low-resource languages and domains. [ABSTRACT FROM AUTHOR]

As an effective way to ensure the economic benefits of enterprises, process operating performance assessment has attracted more and more attention from industry and academia in recent years. In this paper, a stacked performance‐relevant enhanced denoising autoencoder (SPEDAE) network is designed for the operating performance assessment of industrial processes. Compared to the original denoising auto‐encoder (DAE), each performance‐relevant enhanced denoising auto‐encoder (PEDAE) not only reconstructs the input features in the output layer, but also strives to reconstruct the original input data and the performance grade labels simultaneously. Then the SPEDAE is formed by stacking multiple PEDAEs layer by layer. Through this improved training strategy, SPEDAE can avoid accumulated information loss during the deep feature extraction process, improve the robustness of the network, and extract features closely related to the operating performance, thereby better completing the assessment task. The effectiveness of the proposed assessment method is validated on the case of gold cyanide leaching process. Compared with several methods, the proposed SPEDAE has the highest accuracy and reaches 99.85%, which demonstrates its superiority in operating performance assessment. [ABSTRACT FROM AUTHOR]

With the increase in cyber security attacks, organizations tend to use an intrusion detection system (IDS) based on machine learning. Through the years, IDS based on machine learning has shown their effectiveness in protecting one against attacks. Aside from the machine learning nature being a black-box, there is a possibility of adversaries that can mess up the classification model. Using machine learning in critical aspects such as the medical field and intrusion detection system can result in disastrous impacts on organizations if it is vulnerable to adversary attacks. This paper proposes a new defense approach based on denoising auto-encoder (DAE) to protect IDS from adversarial attacks. To verify the efficacy of the proposed defense mechanism in mitigating adversarial attacks, two datasets were used. The experimental results show that the proposed defense mechanism proves validity against four white-box attacks and one black-box attack. The system's accuracy under adversarial attack elevates from around 68% to 90% and 97% under normal conditions on the first dataset. Similarly, on the second dataset, the models' accuracy increases from 64 to 85% under normal conditions and adversarial attacks. [ABSTRACT FROM AUTHOR]

The water beam mark is a common problem in slab heating, which causes quality defects on strip steel. In hot strip rolling, the heating quality of the slab considerably influences the rolling stability and quality of the finished strip. The water beam mark caused by the heating process and equipment is a common defect in the slab heating. A slab water beam imprint has a great influence on the control precision of the rolling force and thickness of the finished strip. Presently, recognizing the water beam mark is difficult and the workload in the industry is heavy. To solve these problems, this study proposed a recognition algorithm of a hot-rolled strip steel water beam mark based on a semisupervised learning model of an improved denoising autoencoder (DAE). Based on the DAE, random noise was added to each layer of the coding layer, a classification layer was added after a hidden layer, and fake labels were added to the training data. Decoding and classification training are conducted simultaneously. These methods result in the model becoming semisupervised. In this study, we extract the temperature difference of the strip temperature data at the outlet of the roughing mill and use it to train the model. Experimental results showed that the algorithm can accurately recognize the water beam mark of strip steel. The classification accuracy of the proposed model is 5.0%–10.0% higher than other mainstream models when the number of tag proportions is small. When the number of tag proportions is large, the accuracy of the proposed model reaches up to 93.8%. According to the result, the production efficiency can be improved using this model.

The advances in single-cell ribonucleic acid sequencing (scRNA-seq) allow researchers to explore cellular heterogeneity and human diseases at cell resolution. Cell clustering is a prerequisite in scRNA-seq analysis since it can recognize cell identities. However, the high dimensionality, noises and significant sparsity of scRNA-seq data have made it a big challenge. Although many methods have emerged, they still fail to fully explore the intrinsic properties of cells and the relationship among cells, which seriously affects the downstream clustering performance. Here, we propose a new deep contrastive clustering algorithm called scDCCA. It integrates a denoising auto-encoder and a dual contrastive learning module into a deep clustering framework to extract valuable features and realize cell clustering. Specifically, to better characterize and learn data representations robustly, scDCCA utilizes a denoising Zero-Inflated Negative Binomial model-based auto-encoder to extract low-dimensional features. Meanwhile, scDCCA incorporates a dual contrastive learning module to capture the pairwise proximity of cells. By increasing the similarities between positive pairs and the differences between negative ones, the contrasts at both the instance and the cluster level help the model learn more discriminative features and achieve better cell segregation. Furthermore, scDCCA joins feature learning with clustering, which realizes representation learning and cell clustering in an end-to-end manner. Experimental results of 14 real datasets validate that scDCCA outperforms eight state-of-the-art methods in terms of accuracy, generalizability, scalability and efficiency. Cell visualization and biological analysis demonstrate that scDCCA significantly improves clustering and facilitates downstream analysis for scRNA-seq data. The code is available at https://github.com/WJ319/scDCCA. [ABSTRACT FROM AUTHOR]

Accurately predicting the Remaining Useful Life (RUL) of a Li-ion battery plays an important role in managing the health and estimating the state of a battery. With the rapid development of electric vehicles, there is an increasing need to develop and improve the techniques for predicting RUL. To predict RUL, we designed a Transformer-based neural network. First, battery capacity data is always full of noise, especially during battery charge/discharge regeneration. To alleviate this problem, we applied a Denoising Auto-Encoder (DAE) to process raw data. Then, to capture temporal information and learn useful features, a reconstructed sequence was fed into a Transformer network. Finally, to bridge denoising and prediction tasks, we combined these two tasks into a unified framework. Results of extensive experiments conducted on two data sets and a comparison with some existing methods show that our proposed method performs better in predicting RUL. Our projects are all open source and are available at https://github.com/XiuzeZhou/RUL.

Ningxia wolfberry is the only wolfberry product with medicinal value in China. However, the nutritional elements, active ingredients, and economic value of the wolfberry vary considerably among different origins in Ningxia. It is difficult to determine the origin of wolfberry by traditional methods due to the same variety, similar origins, and external characteristics. In the study, we have for the first time used a multi-task residual fully convolutional network (MRes-FCN) under Bayesian optimized architecture for imaging from visible-near-infrared (Vis-NIR, 400–1000 nm) and near-infrared (NIR-1700 nm) hyperspectral imaging (HSI) technology to establish a classification model for near geographic origin of Ningxia wolfberries (Zhongning, Guyuan, Tongxin, and Huinong). The denoising auto-encoder (DAE) was used to generate augmented data, then principal component analysis (PCA) was combined with gray level co-occurrence matrix (GLCM) to extract the texture features. Finally, three datasets (HSI, DAE, and texture) were added to the multi-task model. The reshaped data were up-sampled using transposed convolution. After data-sparse processing, the backbone network was imported to train the model. The results showed that the MRes-FCN model exhibited excellent performance, with the accuracies of the full spectrum and optimum characteristic spectrum of 95.54% and 96.43%, respectively. This study has demonstrated that the MRes-FCN model based on Bayesian optimization and DAE data augmentation strategy may be used to identify the near geographical origin of wolfberries. [ABSTRACT FROM AUTHOR]

Unknown cyber-attacks have appeared constantly. Several anomaly detection techniques based on semi-supervised learning have been proposed to detect these unknown cyber-attacks. Among them, the Denoising Auto-Encoder (DAE) scheme performs better than others in accuracy but is not good enough in precision. This paper proposes a novel two-stage deep learning structure for network flow anomaly detection by combining the models of Gate Recurrent Unit (GRU) and DAE. By using supervised anomaly detection with a selection mechanism to assist semi-supervised anomaly detection, the precision and accuracy of the anomaly detection system are improved. In the proposed structure, we first use the GRU model to analyze the network flow and then take the outcome from the Softmax function as a confidence score. When the score is more than or equal to the predefined confidence threshold, the GRU model outputs the flow as a positive result, no matter the flow is classified as normal or abnormal. When the score is less than the confidence threshold, GRU model outputs the flow as a negative result and passes the flow to DAE model for flow classification. DAE then determines a reconstruction error threshold by learning the pattern of normal flows. Accordingly, the flow is normal or abnormal depending on whether it is under or over the reconstruction error threshold. A comparative experiment is performed using NSL-KDD dataset as benchmark. The results revealed that the precision using the proposed scheme is 0.83% better than DAE. The accuracy using the proposed approach is 90.21%, which is better than Random Forest, Naïve Bayes, One-Dimensional Convolutional Neural Network, two-stage Auto-Encoder, etc. In addition, the proposed approach is also applied to the environment of software defined network (SDN). By adopting our approach in SDN environment, the precision and F-measure are significantly improved. [ABSTRACT FROM AUTHOR]

Mounting concerns pertaining to energy efficiency have led to the research of load monitoring. By Non-Intrusive Load Monitoring (NILM), detailed information regarding the electric energy consumed by each appliance per day or per hour can be formed. The accuracy of the previous residential load monitoring approach relies heavily on the data acquisition frequency of the energy meters. It brings high overall cost issues, and furthermore, the differentiating algorithm becomes much more complicated. Based on this, we proposed a novel non-Intrusive residential load disaggregation method that only depends on the regular data acquisition speed of active power measurements. Additionally, this approach brings some novelties to the traditionally used denoising Auto-Encoder (dAE), i.e., the reconfiguration of the overlapping parts of the sliding windows. The median filter is used for the data processing of the overlapping window. Two datasets, i.e., the Reference Energy Disaggregation Dataset (REDD) and TraceBase, are used for test and validation. By numerical testing of the real residential data, it proves that the proposed method is superior to the traditional Factorial Hidden Markov Model (FHMM)-based approach. Furthermore, the proposed method can be used for energy data, disaggregation disregarding the brand and model of each appliance. [ABSTRACT FROM AUTHOR]

In real-world datasets, the number of samples in each class is often imbalanced, which results in the classifier's suboptimal performance. Presently, the imbalanced binary classification approach based on deep learning has achieved good results and gets more attention constantly. In this study, we present a dual encoder (Denoising Auto-Encoder) DAE neural network based on non-dominated sorting genetic algorithm (NSGA-III) and generative adversarial network (GAN) to address the imbalanced binary classification problem. The primary aim of our approach is to increase the separability between the reconstruction error of minority class latent features and the reconstruction error of majority class latent features. For this purpose, we first create a dual encoder DAE network to obtain the reconstruction error of latent features of training data. Second, when training the neural network, we introduced GAN to perform a layer-wise training which can improve the training effect of the model. Third, in order to increase the separability of the reconstruction error of minority class and majority class, we utilize NSGA-III to optimize the parameters of the second encoder. Then, we can obtain a set of non-dominated solutions. Finally, based on the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method, we can get the best solution, which is the most appropriate parameter set of the second encoder to distinguish the minority class and the majority class. The experiment results on both benchmark datasets and a real-world dataset for communication anomaly detection demonstrate the superiority of the proposed approach in imbalanced binary classification problem. [ABSTRACT FROM AUTHOR]

Identification of the type of communication technology and/or modulation scheme based on detected radio signal are challenging problems encountered in a variety of applications including spectrum allocation and radio interference mitigation. They are rendered difficult due to a growing number of emitter types and varied effects of real-world channels upon the radio signal. Existing spectrum monitoring techniques are capable of acquiring massive amounts of radio and real-time spectrum data using compact sensors deployed in a variety of settings. However, state-of-the-art methods that use such data to classify emitter types and detect communication schemes struggle to achieve required levels of accuracy at a computational efficiency that would allow their implementation on low-cost computational platforms. In this paper, we present a learning framework based on an LSTM denoising auto-encoder designed to automatically extract stable and robust features from noisy radio signals, and infer modulation or technology type using the learned features. The algorithm utilizes a compact neural network architecture readily implemented on a low-cost computational platform while exceeding state-of-the-art accuracy. Results on realistic synthetic as well as over-the-air radio data demonstrate that the proposed framework reliably and efficiently classifies received radio signals, often demonstrating superior performance compared to state-of-the-art methods. Source codes are available at https://github.com/WuLoli/LSTMDAE. [ABSTRACT FROM AUTHOR]

The 3D deformable image registration is one of the most challenging tasks in medical image analysis. Due to the large and complex deformation in 3D medical images, many deep neural network based methods have been proposed to improve the image similarity after registration, among which recursive cascading network structure is one of the state-of-the-art. However, most existing works rely on the pixel-level image similarities to achieve anatomical rationality and overlook the global-level resemblance between the two structures. Therefore, the resulting registration is not quite clinically valuable. To this end, in this work, we propose a Progressive Anatomically Constrained deep neural Network (PACN) to incorporate the anatomical priors into a progressive cascading registration network to improve the anatomical plausibility as well as the pixel-level similarity of the registration results. Specifically, an Anatomical Constraint Encoder (ACE) network is proposed to encode the global context of the anatomical segmentations and attached to the dense registration network to form a registration unit. Repeated such units forming a cascading framework progressively warps the moving image toward the fixed one, with the output warped image of one unit as the input of the next unit. In this design, the global anatomical priors along with the pixel-level local information are used to guide the model learning process to produce high quality deformation field. Based on this, we explore two frameworks to investigate their registration effectiveness, one attaches the anatomical constraint encoder (ACE) to every dense registration sub-network and the other one attaches ACE only to the last dense registration unit. We test the two frameworks on benchmarks of three liver image datasets SLIVER, LiTS and LSPIG, and one brain dataset LPBA. Our two frameworks have achieved significantly better results in terms of average Dice score than the state-of-the-art baseline method on three liver datasets and comparable on LPBA when both tested with up to three cascades. [ABSTRACT FROM AUTHOR]

Copyright of China Mechanical Engineering is the property of Editorial Board of China Mechanical Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

To solve the Grammatical Error Correction (GEC) problem , a mapping between a source sequence and a target one is needed, where the two differ only on few spans. For this reason, the attention has been shifted to the non-autoregressive or sequence tagging models. In which, the GEC has been simplified from Seq2Seq to labeling the input tokens with edit commands chosen from a large edit space. Due to this large number of classes and the limitation of the available datasets, the current sequence tagging approaches still have some issues handling a broad range of grammatical errors just by being laser-focused on one single task. To this end, we simplified the GEC further by dividing it into seven related subtasks: Insertion, Deletion, Merge, Substitution, Transformation, Detection, and Correction, with Correction being our primary focus. A distinct classification head is dedicated to each of these subtasks. The novel multi-head and multi-task learning model is proposed to effectively utilize training data and harness the information from related task training signals. To mitigate the limited number of available training samples, a new denoising autoencoder is used to generate a new synthetic dataset to be used for pretraining. Additionally, a new character-level transformation is proposed to enhance the sequence-to-edit function and improve the model's vocabulary coverage. Our single/ensemble model achieves an F0.5 of 74.4/77.0, and 68.6/69.1 on BEA-19 (test) and CoNLL-14 (test) respectively. Moreover, evaluated on JFLEG test set, the GLEU scores are 61.6 and 61.7 for the single and ensemble models, respectively. It mostly outperforms recently published state-of-the-art results by a considerable margin. [Display omitted] • A new GEC model simplifies tagging into subtasks, learned jointly. • A denoising auto-encoder converts vast unlabeled data into artificial labeled one. • A character-level transformation mitigates the effect of the limited vocabulary size. [ABSTRACT FROM AUTHOR]

Deep learning models often need large amounts of training samples (thousands of training samples) to effectively extract hidden patterns in the data, thus achieving better results. However, in the field of brain-related disease, the omics data obtained by using advanced sequencing technology typically have much fewer patient samples (tens to hundreds of samples). Due to the small sample problem, statistical methods and intelligent machine learning methods have been unable to obtain a convergent gene set when prioritizing biomarkers. Furthermore, mathematical models designed for prioritizing biomarkers perform differently on different datasets. However, the architecture of the generative adversarial network (GAN) can address this bottleneck problem. Through the game between the generator and the discriminator, samples with similar distributions to that of samples in the training set can be generated by the generator, and the prediction accuracy and robustness of the discriminator could be significantly improved. Therefore, in this study, we designed a new generative adversarial network model with a denoising auto-encoder (DAE) as the generator and a multilayer perceptron (MLP) as the discriminator. The prediction residual error was backpropagated to the decoder part of the DAE, modifying the captured probability distribution. Based on this model, we further designed a framework to predict disease genes with RNA-seq data. The deep learning model improves the identification accuracy of disease genes over the-state-of-the-art approaches. An analysis of the experimental results has uncovered new disease-related genes and disease-associated pathways in the brain, which in turn have provided insight into the molecular mechanisms underlying disease phenotypes.

Online learning is an important technical means for sketching massive real-time and high-speed data. Although this direction has attracted intensive attention, most of the literature in this area ignore the following three issues: (1) they think little of the underlying abstract hierarchical latent information existing in examples, even if extracting these abstract hierarchical latent representations is useful to better predict the class labels of examples; (2) the idea of preassigned model on unseen datapoints is not suitable for modeling streaming data with evolving probability distribution. This challenge is referred as "model flexibility". And so, with this in minds, the online deep learning model we need to design should have a variable underlying structure; (3) moreover, it is of utmost importance to fusion these abstract hierarchical latent representations to achieve better classification performance, and we should give different weights to different levels of implicit representation information when dealing with the data streaming where the data distribution changes. To address these issues, we propose a two-phase Online Deep Learning based on Auto-Encoder (ODLAE). Based on auto-encoder, considering reconstruction loss, we extract abstract hierarchical latent representations of instances; Based on predictive loss, we devise two fusion strategies: the output-level fusion strategy, which is obtained by fusing the classification results of encoder's each hidden layer; and feature-level fusion strategy, which is leveraged self-attention mechanism to fusion the every hidden layer's output. Finally, in order to improve the robustness of the algorithm, we also try to utilize the denoising auto-encoder to yield hierarchical latent representations. Experimental results on different datasets are presented to verify the validity of our proposed algorithm (ODLAE) outperforms several baselines. [ABSTRACT FROM AUTHOR]

Ningxia wolfberry is the only wolfberry product with medicinal value in China. However, the nutritional elements, active ingredients, and economic value of the wolfberry vary considerably among different origins in Ningxia. It is difficult to determine the origin of wolfberry by traditional methods due to the same variety, similar origins, and external characteristics. In the study, we have for the first time used a multi-task residual fully convolutional network (MRes-FCN) under Bayesian optimized architecture for imaging from visible-near-infrared (Vis-NIR, 400–1000 nm) and near-infrared (NIR-1700 nm) hyperspectral imaging (HSI) technology to establish a classification model for near geographic origin of Ningxia wolfberries (Zhongning, Guyuan, Tongxin, and Huinong). The denoising auto-encoder (DAE) was used to generate augmented data, then principal component analysis (PCA) was combined with gray level co-occurrence matrix (GLCM) to extract the texture features. Finally, three datasets (HSI, DAE, and texture) were added to the multi-task model. The reshaped data were up-sampled using transposed convolution. After data-sparse processing, the backbone network was imported to train the model. The results showed that the MRes-FCN model exhibited excellent performance, with the accuracies of the full spectrum and optimum characteristic spectrum of 95.54% and 96.43%, respectively. This study has demonstrated that the MRes-FCN model based on Bayesian optimization and DAE data augmentation strategy may be used to identify the near geographical origin of wolfberries.

Throat microphones (TM) are a type of skin-attached non-acoustic sensors, which are robust to environmental noise but carry a lower signal bandwidth characterization than the traditional close-talk microphones (CM). Attaining high-performance phoneme recognition is a challenging task when the training data from a degrading channel, such as TM, is limited. In this paper, we address this challenge for the TM speech recordings using a transfer learning approach based on the stacked denoising auto-encoders (SDA). The proposed transfer learning approach defines an SDA-based domain adaptation framework to map the source domain CM representations and the target domain TM representations into a common latent space, where the mismatch across TM and CM is eliminated to better train an acoustic model and to improve the TM phoneme recognition. For the phoneme recognition task, we use the convolutional neural network (CNN) and the hidden Markov model (HMM) based CNN/HMM hybrid system, which delivers better acoustic modeling performance compared to the conventional Gaussian mixture model (GMM) based models. In the experimental evaluations, we observed more than 12% relative phoneme error rate (PER) improvement for the TM recordings with the proposed transfer learning approach compared to baseline performances. • High-performance phoneme recognition of throat microphone is a challenging task. • Data is scarce for throat microphone speech compared to close-talk microphones. • Transfer learning from close-talk to target throat microphone eliminates mismatch. • Transfer learning as well performs data augmentation from source to target domain. • Proposed learning scheme attains significant phoneme recognition improvements. [ABSTRACT FROM AUTHOR]

With the development of big data, artificial intelligence has provided many intelligent solutions to urban life. For instance, an image-based intelligent technology, such as image classification of diseases, is widely used in daily life. However, the image in real life is mostly unlabeled, so the performance of many image-based intelligent models shows limitations. Therefore, how to use a large amount of unlabeled image data to build an efficient and high-quality model for better urban life has been an urgent research topic. In this paper, we propose an unsupervised image feature extraction method that is referred to as a stacked multi-granularity convolution denoising auto-encoder (SMGCDAE). The algorithm is based on a convolutional neural network (CNN), yet it introduces a multi-granularity kernel. This approach resolved issues with image unicity by extracting a diverse category of high-level features. In addition, the denoising auto-encoder ensures stability and improves the classification accuracy by extracting more robust features. The algorithm was assessed using three image benchmark datasets and a series of meningitis images, achieving higher average accuracy than other methods. These results suggest that the algorithm is capable of extracting more discriminative high-level features and thus offers superior performance compared with the existing methodologies.

Data loss for electronic noses may occur because of the sensor’s installation environment or from electrical disturbances. As a result, electronic noses may experience difficulties when identifying gases. This paper proposes two deep neural network-based functions for identifying gases. First, a denoising auto-encoder based on the corruption reconstruction method is proposed for electronic nose data to solve this problem. Second, a convolutional neural network-based gas-classifying model is proposed. Although the electronic nose data are highly discriminative, they are sensitive to the corruption of information; hence, they require an efficient restoration method for practical use. From the experiments we demonstrate that the proposed denoising auto-encoder provides a strong restoration capability, and the convolutional neural network-based classifier successfully discriminates the gas data samples with a classification rate over 95% even when the data loss is 50%. [ABSTRACT FROM AUTHOR]

In the absence of a priori knowledge, manual feature selection is too blind to find the sensitive features which can effectively classify the different fault features. And it is difficult to obtain a large number of typical fault samples in practice to train the intelligent classifier. A novel intelligent fault diagnosis method based on feature selection and deep learning is proposed for rotating machine mechanical in the paper. In this method, the deep neural network is not only used for feature extraction but also for fault diagnosis. First, the deep neural network 1 is used to extract feature from the spectral signal of the original signal. In addition, the original vibration signal is decomposed to a series of intrinsic mode function components by empirical mode decomposition, and the statistical features of each intrinsic mode function component are extracted by the deep neural network 2 in time domain and frequency domain. Second, the extraction features of the original signal spectrum and the extraction features of each intrinsic mode function component are evaluated, respectively. After features evaluation, the selected sensitive features are combined together to construct a joint feature. Finally, the joint feature is put into the deep neural network 3 to realize the automatic recognition of different fault states of rotating machinery. The experimental results show that the method proposed in this paper which integrated time-domain, frequency-domain statistical characteristics, empirical mode decomposition, feature selection, and deep learning methods can obtain the fault information in detail and can select sensitive features from a large number of fault features. The method can reduce the network size, improve the mechanical fault diagnosis classification accuracy, and has strong robustness. [ABSTRACT FROM AUTHOR]

Gene expression analysis plays a significant role for providing molecular insights in cancer. Various genetic and epigenetic factors (being dealt under multi-omics) affect gene expression giving rise to cancer phenotypes. A recent growth in understanding of multi-omics seems to provide a resource for integration in interdisciplinary biology since they altogether can draw the comprehensive picture of an organism's developmental and disease biology in cancers. Such large scale multi-omics data can be obtained from public consortium like The Cancer Genome Atlas (TCGA) and several other platforms. Integrating these multi-omics data from varied platforms is still challenging due to high noise and sensitivity of the platforms used. Currently, a robust integrative predictive model to estimate gene expression from these genetic and epigenetic data is lacking. In this study, we have developed a deep learning-based predictive model using Deep Denoising Auto-encoder (DDAE) and Multi-layer Perceptron (MLP) that can quantitatively capture how genetic and epigenetic alterations correlate with directionality of gene expression for liver hepatocellular carcinoma (LIHC). The DDAE used in the study has been trained to extract significant features from the input omics data to estimate the gene expression. These features have then been used for back-propagation learning by the multilayer perceptron for the task of regression and classification. We have benchmarked the proposed model against state-of-the-art regression models. Finally, the deep learning-based integration model has been evaluated for its disease classification capability, where an accuracy of 95.1% has been obtained. • Integration of genome, epigenome, transcriptome under deep learning (DL) framework. • Estimation of gene expression pattern from CNV and DNA methylation data. • Effectiveness of DL method demonstrated on TCGA liver hepatocellular carcinoma data. • Superior performance of DL method over others in estimating gene expression pattern. • DL based extracted features depicted effective disease classification capability. [ABSTRACT FROM AUTHOR]

The bearing is the core part of the electric spindle, and the bearing failure directly affects the normal operation of the electric spindle. When the electric spindle breaks down, the mechanical system can not operate normally, which leads to great losses. In order to detect bearing faults, many traditional intelligent fault detection methods are proposed, and the fault category can not be accurately determined due to the existence of noise components in the signal. In this paper, a new noise reduction algorithm- Sparse Denoising Auto-Encode (SDAE) is proposed. The Denoising Auto-Encoder can learn the noise factors and extract the succinct expressions from raw data automatically, and sparsity is integrated on the basis of Denoising Auto-Encoder to improve the generalization of feature expression. More effective feature expressions are extracted to train Convolutional Neural Network (CNN), and the Adam optimization algorithm is used to fine-tune CNN to improve the accuracy of fault diagnosis. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Dalian University of Technology / Dalian Ligong Daxue Xuebao is the property of Journal of Dalian University of Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Control Theory & Applications / Kongzhi Lilun Yu Yinyong is the property of Editorial Department of Control Theory & Applications and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Wi-Fi based localization has become one of the most practical methods for mobile users in location-based services. However, due to the interference of multipath and high-dimensional sparseness of fingerprint data, with the localization system based on received signal strength (RSS), is hard to obtain high accuracy. In this paper, we propose a novel indoor positioning method, named JLGBMLoc (Joint denoising auto-encoder with LightGBM Localization). Firstly, because the noise and outliers may influence the dimensionality reduction on high-dimensional sparseness fingerprint data, we propose a novel feature extraction algorithm—named joint denoising auto-encoder (JDAE)—which reconstructs the sparseness fingerprint data for a better feature representation and restores the fingerprint data. Then, the LightGBM is introduced to the Wi-Fi localization by scattering the processed fingerprint data to histogram, and dividing the decision tree under leaf-wise algorithm with depth limitation. At last, we evaluated the proposed JLGBMLoc on the UJIIndoorLoc dataset and the Tampere dataset, the experimental results show that the proposed model increases the positioning accuracy dramatically compared with other existing methods.

In this paper, we address the poor generalization ability problem of traditional auto-encoder on noise data, and propose a Lossless-constraint Denoising (LD) method, which can enhance the anti-noise ability and robustness of auto-encoders. We respectively utilize the denoising capability of Denoising Auto-encoder (DAE) and Sparse Auto-encoder (SAE), design two auto-encoders of better noise immunity: Lossless-constraint Denoising Auto-encoder (LDAE) and Lossless-constraint Denoising Sparse Auto-encoder (LDSAE). [ABSTRACT FROM AUTHOR]

Ensemble clustering aims to fuse several diverse basic partitions into a consensus one, which has been widely recognized as a promising tool to discover novel clusters and deliver robust partitions, while representation learning with deep structure shows appealing performance in unsupervised feature pre-treatment. In the literature, it has been empirically found that with the increasing number of basic partitions, ensemble clustering gets better performance and lower variances, yet the best number of basic partitions for a given data set is a pending problem. In light of this, we propose the Infinite Ensemble Clustering (IEC), which incorporates marginalized denoising auto-encoder with dropout noises to generate the expectation representation for infinite basic partitions. Generally speaking, a set of basic partitions is firstly generated from the data. Then by converting the basic partitions to the 1-of-K codings, we link the marginalized denoising auto-encoder to the infinite basic partition representation. Finally, we follow the layer-wise training procedure and feed the concatenated deep features to K-means for final clustering. According to different types of marginalized auto-encoders, the linear and non-linear versions of IEC are proposed. Extensive experiments on diverse vision data sets with different levels of visual descriptors demonstrate the superior performance of IEC compared to the state-of-the-art ensemble clustering and deep clustering methods. Moreover, we evaluate the performance of IEC in the application of pan-omics gene expression analysis application via survival analysis. [ABSTRACT FROM AUTHOR]

With the rapidly development of natural language processing science, sentiment analysis as one of its important branches are widely used in social network platform. Especially the micro-blogging which are wide dissemination and contains a wealth of information on the emotional and highly favored by scholars. Due to the micro-blogging widely spread and contains rich human emotional information, it quickly become the research object of the Chinese and foreign scholars. To analysis the expression of human emotion in the micro-blogging and even digging its inherent sentiment, this paper made a further on the de-noising auto-encoder( DAE) model and explored a new method to improve it. The characteristics of DAE was to achieve the reduction of the original input under the interference of noise,while the advantage of the new model were take into account the diversity and complexity of noise,and to strengthen the resilience of the original features of the model through the deep learning training,then the result was overcame the original input noise which could not prediction. At last,by separately using SVM, DAE model and the improved model made sentiment analysis experiment, comparing the classification results indicate that the improved model(IDAE) are more accurate in micro-blogging sentiment analysis. Moreover, its anti-interference ability and robustness has improved. [ABSTRACT FROM AUTHOR]

Conventional learning algorithm assumes that the training data and test data share a common distribution. However, this assumption will greatly hinder the practical application of the learned model for cross-domain data analysis in multimedia. To deal with this issue, transfer learning based technology should be adopted. As a typical version of transfer learning, domain adaption has been extensively studied recently due to its theoretical value and practical interest. In this article, we propose a boosted multifeature learning (BMFL) approach to iteratively learn multiple representations within a boosting procedure for unsupervised domain adaption. The proposed BMFL method has a number of properties. (1) It reuses all instances with different weights assigned by the previous boosting iteration and avoids discarding labeled instances as in conventional methods. (2) It models the instance weight distribution effectively by considering the classification error and the domain similarity, which facilitates learning new feature representation to correct the previously misclassified instances. (3) It learns multiple different feature representations to effectively bridge the source and target domains. We evaluate the BMFL by comparing its performance on three applications: image classification, sentiment classification and spam filtering. Extensive experimental results demonstrate that the proposed BMFL algorithm performs favorably against state-of-the-art domain adaption methods. [ABSTRACT FROM AUTHOR]

Sensors deployed all over the buildings are nowadays collecting a large amount of data, such as the Indoor Air Quality (IAQ) data which can provide valuable suggestions on improving indoor environments and energy consumption strategies. However, as treated as Multivariate Time Series (MTS), IAQ data often contain missing values that severely limit further analysis on them. Unfortunately, most of the existing methods fail to handle a couple of technical issues due to the complexity of MTS data, such as data distribution approximation, removing the redundancy, and so on. In this paper, we formulate the IAQ missing data imputation problem and propose an Inverse Mapping Generative Adversarial Network (IM-GAN) to tackle that problem. IM-GAN takes advantage of Bi-directional Recurrent Neural Network (BRNN), Denoising Auto-Encoder (DAE), and Generative Adversarial Network (GAN) to overcome the aforementioned technical issues. To validate the effectiveness of our proposed IM-GAN, we conduct comprehensive experiments on two public IAQ datasets GAMS and Gainesville. Results show that our IM-GAN achieves the new state-of-the-art performance in accurately estimating missing values in indoor air quality time series data, with the average performance of 0.1566 and 0.0789 in terms of Mean Relative Error , and 17.2884 and 2.7434 in terms of Mean Absolute Error on GAMS and Gainesville respectively at different missing rates. Our ablation study and visualization also validate that IM-GAN effectively overcomes the aforementioned technical issues by capturing data distribution, eliminating network saturation, and so on for IAQ data imputation. • Four key technical issues are identified for indoor air quality data imputation. • A novel model named IM-GAN is proposed to impute incomplete indoor air quality data. • IM-GAN significantly outperforms the baselines in comprehensive experiments. [ABSTRACT FROM AUTHOR]

With the frequent occurrence of safety incidents in cyber–physical systems (CPSs), great significance has been attached to the study of defense schemes against cyber–physical attacks. In this paper, an integrated data-driven defense scheme is proposed, which can sensitively detect data integrity attacks such as false data injection (FDI) attacks, denial-of-service (DoS) attacks, and replay attacks, and ensures secure transmission against eavesdropping attacks. Specifically, a novel deep learning model is designed so that both the online detection task and the encryption/decryption task can be completed under the same framework. The main idea is inspired by denoising auto-encoders whereas necessary changes are made to adapt to the challenges in the context of CPS attacks, and in light of this, the proposed approach is called modified denoising auto-encoder (MDAE). Unlike supervised classifier-based detectors, the proposed detector can retain sensitivity to unknown attacks because it is trained to learn the normal operation behavior. Moreover, to improve the detectability of the DoS and replay attacks on all data, the check code is designed. Encrypting the transmitted data through nonlinear mapping is achieved using the same MDAE, which prevents the attackers from recording useful information. Benefiting from the fact that the dimension of the variables is reduced after encryption, the transmission traffic can be saved. Simulation results on the measurement data instances generated by the IEEE 118-bus system validate the encryption effects and detection accuracy of the proposed scheme and show the superiority by comparison study. • A modified denoising autoencoder-based detector with check code is designed. • Detecting different types of data integrity attacks with unknown patterns. • An encryption scheme to prevent attackers from getting useful information. • Reducing network transmission traffic and errors caused by noise. [ABSTRACT FROM AUTHOR]

Wi-Fi based localization has become one of the most practical methods for mobile users in location-based services. However, due to the interference of multipath and high-dimensional sparseness of fingerprint data, with the localization system based on received signal strength (RSS), is hard to obtain high accuracy. In this paper, we propose a novel indoor positioning method, named JLGBMLoc (Joint denoising auto-encoder with LightGBM Localization). Firstly, because the noise and outliers may influence the dimensionality reduction on high-dimensional sparseness fingerprint data, we propose a novel feature extraction algorithm—named joint denoising auto-encoder (JDAE)—which reconstructs the sparseness fingerprint data for a better feature representation and restores the fingerprint data. Then, the LightGBM is introduced to the Wi-Fi localization by scattering the processed fingerprint data to histogram, and dividing the decision tree under leaf-wise algorithm with depth limitation. At last, we evaluated the proposed JLGBMLoc on the UJIIndoorLoc dataset and the Tampere dataset, the experimental results show that the proposed model increases the positioning accuracy dramatically compared with other existing methods. [ABSTRACT FROM AUTHOR]