Search

Copyright of Urban Mass Transit is the property of Urban Mass Transit 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.)

Objective When using convolutional neural networks (CNNs) for the state detection of catenary system cotter pins, the imbalance between positive and negative samples leads to a low detection rate of missing pins in the network model. Thus, a single-stage detection network is employed for multi-level precise localization of cotter pins, the state features of the cotter pins are reconstructed in combination with SDAE (stacked denoising autoencoder), thereby achieving efficient detection of missing cotter pins. Method First, a single-stage localization detection network is employed for the regression of cotter pin positions, and the localization results are used as input for SDAE. Different levels of depth noise are introduced into various structural layers of the denoising autoencoder network. By minimizing reconstruction errors, the method enhances semantic understanding of the cotter pin localized images, allowing for accurate assessment of their condition. Additionally, due to the constraints on the size of the localized images, the computational load of the SDAE is relatively low, same with the network time complexity. Result & Conclusion Extensive experimental results demonstrate that the SDAE, based on YOLO v5 algorithm, can accurately detect missing cotter pins at various locations within the catenary system.

To improve the accuracy of terminal area meteorological scene pattern recognition, this study adopts a clustering model based on Stacked Denoising Autoencoder. Noise is added to the input layer, and a three-layer autoencoder is constructed for greedy layer-wise training. The reduced-dimensional features are used as inputs for clustering to achieve meteorological scene pattern recognition. The method is validated using one year of meteorological data from Tianjin Binhai International Airport. Traditional similarity distance measures such as Euclidean distance, Hamming distance, and Manhattan distance are used with both K-medoids and FCM clustering methods. The results show that the similarity measure based on SDAE performs the best in both K-medoids and FCM clustering, with a difference rate of 22.4%, 12%, 17.7%, and 24.8%, 10.7%, 11.8% compared to other similarity measures, respectively. It also has the shortest computation time, demonstrating that the SDAE-based measure and clustering achieve the best performance. Ultimately, eight meteorological scenes are identified with clear and distinct classifications. [ABSTRACT FROM AUTHOR]

Copyright of Railway Construction Technology is the property of Railway Construction Technology 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 era of big data, the exponentially increasing data volume and emerging technical tools have put forward new requirements for enterprise information management. Therefore, it is of great significance to enhance the core competitiveness of enterprises to explore how big data can empower the innovation of enterprise information management. Intelligent transportation system combines a variety of technologies and applies them to a large-scale transportation management system, so as to make a reasonable dispatch of traffic conditions. Aiming at the problem of the relatively low accuracy of bus passenger flow forecasting with the existing models, a short-term passenger flow prediction model combining Stacked Denoising Auto Encoder (SDAE) and improved bidirectional Long-short Term Memory network (Bi-LSTM) is proposed. First, the SDAE model is used to fill in the missing bus passenger flow data, the characteristics of the bus passenger flow data are effectively utilized, and the data with rich information is used to predict the missing values with high accuracy. Second, Bi-LSTM model combined with attention mechanism is used for short-term bus passenger flow prediction. Considering that the data sequence of bus passenger flow is relatively long and there is a two-way information flow, the BiLSTM neural network is used for prediction tasks, and the influence of key factors is highlighted through attention weights to mine the internal laws of passenger flow data. The experimental results show that the proposed method achieves the lowest prediction error among all the comparison methods in the task of short-term bus passenger flow prediction on the public transportation dataset, with MAE, MRE, and RMSE values of 6.014, 0.052, and 9.874, respectively. These findings confirmed the effectiveness of the new model in the passenger flow prediction field. [ABSTRACT FROM AUTHOR]

Early bird One of the most common neurological conditions caused by gradual brain degeneration is Parkinson's disease (PD). Although this neurological condition has no known treatment, early detection and therapy can help patients improve their quality of life. An essential patient's health record is made of medical images used to control, manage, and treat diseases. However, in computerbased diagnostics, disease classification is a difficult task because of the time consumption and high rate of false positive marks. To overcome this problem, this paper introduces a stacked denoising Autoencoder (SDA) for Parkinson's disease classification. In preprocessing, noise is reduced and important information is retained, resulting in increased performance and data augmentation is performed to avoid overfitting issues and increase the size of the dataset. The main aim of this paper is to derive an optimal feature selection design for an effective Parkinson's disease classification. Improved Pigeon-Inspired Optimization (IPIO) algorithm is introduced to enhance the performance of the classifier. Thus, the classification result improved by the optimal features and also increased the sensitivity, accuracy, and specificity in the medical image diagnosis. The proposed scheme is implemented in PYTHON and compared with traditional feature selection models and other classification approaches. The efficacy of the performances is evaluated using a Parkinson's Progression Markers Initiative (PPMI) dataset. The integration of the stacked denoising autoencoder and Improved pigeon inspired optimization method produced the greatest results, with 99.17% accuracy, 98.74% sensitivity, and 98.96% speci- ficity. Furthermore, our finding outperforms the most recent research in the field. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Chongqing University of Technology (Natural Science) is the property of Chongqing 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.)

Visual saliency detection is usually regarded as an image pre-processing method to predict and locate the position and shape of saliency regions. However, many existing saliency detection methods can only obtain the local or even incorrect position and shape of saliency regions, resulting in incomplete detection and segmentation of the salient target region. In order to solve this problem, a visual saliency detection method based on scale invariant feature and stacked denoising autoencoder is proposed. Firstly, the deep belief network would be pretrained to initialize the parameters of stacked denoising autoencoder network. Secondly, different from traditional features, scale invariant feature is not limited to the size, resolution, and content of original images. At the same time, it can help the network to restore important features of original images more accurately in multi-scale space. So, scale invariant feature is adopted to design the loss function of the network to complete self-training and update the parameters. Finally, the difference between the final reconstructed image obtained by stacked denoising autoencoder and the original is regarded as the final saliency map. In the experiment, we test the performance of the proposed method in both saliency prediction and saliency object segmentation. The experimental results show that the proposed method has good ability in saliency prediction and has the best performance in saliency object segmentation than other comparison saliency prediction methods and saliency object detection methods. [ABSTRACT FROM AUTHOR]

Traffic sensing system requires to periodically collect spatial–temporal traffic data distributed among road networks, which results in overhigh bandwidth consumption and storage cost in a large-scale road network. Several compressive sensing-based algorithms are proposed to reconstruct missing traffic data with limited traffic observation. However, there still exist great challenges to be addressed. First, these existing algorithms are always iteration-based, whose time complexity will explosively increase with the growth of network scale. Furthermore, these algorithms have to be re-executed even if only a small fraction of data changes, which is not suitable for dynamic traffic environments. To overcome these issues, we investigate a novel service architecture of traffic sensing based on mobile edge computing where collected data is pre-processed at the edge node and reconstructed at cloud servers, respectively. On this basis, we formulate the problem of Missing Traffic Data reconstruction (MTDR), which aims at maximizing data reconstruction accuracy within limited observation data. Further, we develop a deep-learning-based algorithm called stacked denoising autoencoder for MTDR (SDAE-MTDR), where three denoising autoencoders are trained in order and then stacked together for parameter fine-tuning based on cross-entropy-based loss function. Finally, we conduct comprehensive performance evaluation based on realistic vehicular traces and the simulation results demonstrate the superiority of the proposed algorithm compared with competitive solutions. [ABSTRACT FROM AUTHOR]

Network embedding’s goal is to learn the low-dimensional node feature representation in the network. The learned features are used in various network analysis tasks, such as node classification, link prediction, community detection and recommendation, etc. The existing network embedding methods do not make full use of high-order structure information in social networks. Moreover, the correlation between structure information and node attribute information is not considered. The effect of these methods applied in the social network is not ideal. A social network embedding method combining loop-free path and attributes network embedding (LFNE) is proposed to solve these problems. The high-order structural similarity of nodes is calculated first based on the loop-free path between nodes to eliminate the influence of loop path and large-degree nodes on node structure similarity. This algorithm makes the network embedding method better integrate the high-order social network structure information. Then the node attributes similarity is calculated by combining the loop-free path similarity measurement index between nodes, and the correlation between social network structure information and attribute information is fully utilized to eliminate the noise in attribute information. Finally, the node structure similarity and attribute similarity are fused and applied to learning the low-dimensional feature representation of nodes in the stacked denoising autoencoder. Comparison of experiments with representative algorithms in recent years on three social network datasets shows that the LFNE algorithm can achieve relatively significant results in node classification and link prediction with better network embedding performance.

Community question answering (CQA), with its flexible user interaction characteristics, is gradually becoming a new knowledge-sharing platform that allows people to acquire knowledge and share experiences. The number of questions is rapidly increasing with the open registration of communities and the massive influx of users, which makes it impossible to match many questions to suitable question answering experts (noted as experts) in a timely manner. Therefore, it is of great importance to perform expert recommendation in CQA. Existing expert recommendation algorithms only use data from a single platform, which is not ideal for new CQA platforms with sparse historical interaction and a small number of questions and users. Considering that many mature CQA platforms (source platforms) have rich historical interaction data and a large amount of questions and experts, this paper will fully mine the information and transfer it to new platforms with sparse data (target platform), which can effectively alleviate the data sparsity problem. However, the feature composition of questions and experts in different platforms is inconsistent, so the data from the source platform cannot be directly transferred for training in the target platform. Therefore, this paper proposes feature-alignment-based cross-platform question answering expert recommendation (FA-CPQAER), which can align expert and question features while transferring data. First, we use the rating predictor composed by the BP network for expert recommendation within the domains, and then the feature matching of questions and experts between two domains by similarity calculation is achieved for the purpose of using the information in the source platform to assist expert recommendation in the target platform. Meanwhile, we train a stacked denoising autoencoder (SDAE) in both domains, which can map user and question features to the same dimension and align the data distributions. Extensive experiments are conducted on two real CQA datasets, Toutiao and Zhihu datasets, and the results show that compared to the other advanced expert recommendation algorithms, this paper's method achieves better results in the evaluation metrics of MAE, RMSE, Accuracy, and Recall, which fully demonstrates the effectiveness of the method in this paper to solve the data sparsity problem in expert recommendation. [ABSTRACT FROM AUTHOR]

In the recommendation systems, the user feedback data (e.g., user-item rating and the social information) is usually sparse, discrete and full of noise. However, most of existing methods use it to make recommendations directly, which leads to the reduction of their recommendation accuracy and quality ultimately. To address this problem, this paper proposes an Enhanced Autoencoder Framework with knowledge distillation (EAF-SR) for learning robust information from user feedback data, where it consists of three parts: Pre-training, knowledge distillation layers and Re-training. Specifically, Pre-training is proposed to generate the soft targets by using the stacked denoising autoencoder (SDAE) from user feedbacks and trust information, which aims to reduce data noise. Then, knowledge distillation layers are developed to learn the robust information from the generated soft targets. Finally, Re-training network combined with Pre-training is designed to make recommendations for each user. Extensive experiments done on three real-world datasets (e.g., Flixster, Epinions and Douban) show that the proposed EAF-SR is superior to the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Many recent data-driven studies have used sensor profile data for prognostics and health management (PHM). However, existing data-driven PHM techniques are vulnerable to three types of uncertainty: sensor noise inherent to the sensor profile data, uncertainty regarding the current health status diagnosis caused by monitoring a single health index (HI), and uncertainty in predicting the remaining useful life (RUL), which is affected by unpredictable changes in system operating conditions and the future external environment. This study proposes a deep conditional health index extraction network (DCHIEN) for PHM to effectively manage these three types of uncertainty. DCHIEN is a model that combines a stacked denoising autoencoder that extracts high-level features robust to sensor noise with a feed-forward neural network that produces an HI based on user-defined monitoring conditions. This approach supports system health monitoring using the conditional HI, as well as prognostics using RUL interval predictions. Extensive experiments were conducted using NASA's turbofan engine degradation dataset. The results show that the proposed method achieves a superior RUL prediction performance compared to state-of-the-art methods and that uncertainties can be effectively managed. [ABSTRACT FROM AUTHOR]

The effective utilisation of monitoring data of the coal mine is the core of realising intelligent mine. The complex and challenging underground environment, coupled with unstable sensors, can result in "dirty" data in monitoring information. A reliable data cleaning method is necessary to figure out how to extract high-quality information from large monitoring data sets while minimising data redundancy. Based on this, a cleaning method for sensor monitoring data based on stacked denoising autoencoders (SDAE) is proposed. The sample data of the ventilation system under normal conditions are trained by the SDAE algorithm and the upper limit of reconstruction errors is obtained by Kernel density estimation (KDE). The Apriori algorithm is used to study the correlation between monitoring data time series. By comparing reconstruction errors and error duration of test data with the upper limit of reconstruction error and tolerance time, cooperating with the correlation rule, the "dirty" data is resolved. The method is tested in the Dongshan coal mine. The experimental results show that the proposed method can not only identify the dirty data but retain the faulty information. The research provides effective basic data for fault diagnosis and disaster warning. [ABSTRACT FROM AUTHOR]

[Display omitted] • Grape shelf-life was successfully estimated by HSI technique and deep learning method. • PCA and LPP were studied and compared to obtain conventional features for classification. • Deep spectral features in Vis/NIR hyperspectral image of grape were extracted by SDAE. • Proposed deep learning models have higher precision than SVM, 1D CNN and ResNet models. Grape shelf-life estimation is a substantial challenge for the grape industry. The objective of this study is to investigate the potential of grape shelf-life estimation using HSI technique and a deep learning algorithm. The visible and near-infrared (400.68–1001.61 nm) hyperspectral reflectance images data of grape samples was acquired and preprocessed with different spectral preprocessing methods. Additionally, a stacked denoising autoencoder (SDAE)-based deep learning algorithm was developed to extract deep features from pixel-level hyperspectral data of grapes, and then these features were used as inputs to establish support vector machine (SVM) models for estimating grape shelf-life. Furthermore, SVM, one-dimensional convolutional neural network (1D CNN) and long short-term memory (LSTM) models were used as traditional machine learning and end to end models for comparison. The results demonstrated that the SDAE-SVM model achieved reasonable recognition accuracy of 100 % and 98.125 % for the shelf-life of grapes in the training and test sets, respectively. The overall results suggested that SDAE-based deep learning method can be used as a powerful tool to deal with large-scale hyperspectral data as well as this research confirms the feasibility of non-destructive estimation for grapes shelf-life by the combination of HSI technique and deep learning method, which would provide a valuable guidance for shelf-life estimation of other postharvest fruit. [ABSTRACT FROM AUTHOR]

To quickly respond to variations in the state of network load demand, a solution using data-driven techniques to predict optimal power flow (OPF) has emerged in recent years. However, most of the existing methods are highly dependent on large data volumes. This limits their application on the newly established or expanded systems. In this regard, this work proposes a sample-efficient OPF learning method to maximize the utilization of limited samples. By decomposing the OPF task before knowledge distillation, deep learning complexity is reduced. Thereafter, knowledge distillation is used to integrate decoupled tasks and improve accuracy in low-data setups. Unsupervised pre-training is introduced to alleviate the demand for labeled data. Additionally, the focal loss function and teacher annealing strategy are adopted to achieve higher accuracy without extra samples. Numerical tests on different systems corroborate the advanced accuracy and training speed over other training methods, especially on fewer-sample occasions.

Copyright of Journal of Frontiers of Computer Science & Technology is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. 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.)

Summary: Intrusion detection is essential to prevent damage to computer systems. However, in recent years, with the development of the network, many complex attack types have appeared, and it has become increasingly difficult to obtain high detection rates and low false alarm rates. In addition, traditional heavily hand‐crafted evaluation datasets for network intrusion detection have not been practical. This article proposes an intrusion detection method based on hierarchical feature learning, which can automatically learn traffic features. The method first learns the byte‐level features of network traffic through one‐dimensional convolutional neural networks and then learns session‐level features using stacked denoising autoencoder. The experiment analyzed the model structure and compared it with other methods. Experiments prove that the method in this article has high accuracy and low false alarm rate. [ABSTRACT FROM AUTHOR]

Abstract Background Plant long non-coding RNAs (lncRNAs) play vital roles in many biological processes mainly through interactions with RNA-binding protein (RBP). To understand the function of lncRNAs, a fundamental method is to identify which types of proteins interact with the lncRNAs. However, the models or rules of interactions are a major challenge when calculating and estimating the types of RBP. Results In this study, we propose an ensemble deep learning model to predict plant lncRNA-protein interactions using stacked denoising autoencoder and convolutional neural network based on sequence and structural information, named PRPI-SC. PRPI-SC predicts interactions between lncRNAs and proteins based on the k-mer features of RNAs and proteins. Experiments proved good results on Arabidopsis thaliana and Zea mays datasets (ATH948 and ZEA22133). The accuracy rates of ATH948 and ZEA22133 datasets were 88.9% and 82.6%, respectively. PRPI-SC also performed well on some public RNA protein interaction datasets. Conclusions PRPI-SC accurately predicts the interaction between plant lncRNA and protein, which plays a guiding role in studying the function and expression of plant lncRNA. At the same time, PRPI-SC has a strong generalization ability and good prediction effect for non-plant data.

Community question answering (CQA), with its flexible user interaction characteristics, is gradually becoming a new knowledge-sharing platform that allows people to acquire knowledge and share experiences. The number of questions is rapidly increasing with the open registration of communities and the massive influx of users, which makes it impossible to match many questions to suitable question answering experts (noted as experts) in a timely manner. Therefore, it is of great importance to perform expert recommendation in CQA. Existing expert recommendation algorithms only use data from a single platform, which is not ideal for new CQA platforms with sparse historical interaction and a small number of questions and users. Considering that many mature CQA platforms (source platforms) have rich historical interaction data and a large amount of questions and experts, this paper will fully mine the information and transfer it to new platforms with sparse data (target platform), which can effectively alleviate the data sparsity problem. However, the feature composition of questions and experts in different platforms is inconsistent, so the data from the source platform cannot be directly transferred for training in the target platform. Therefore, this paper proposes feature-alignment-based cross-platform question answering expert recommendation (FA-CPQAER), which can align expert and question features while transferring data. First, we use the rating predictor composed by the BP network for expert recommendation within the domains, and then the feature matching of questions and experts between two domains by similarity calculation is achieved for the purpose of using the information in the source platform to assist expert recommendation in the target platform. Meanwhile, we train a stacked denoising autoencoder (SDAE) in both domains, which can map user and question features to the same dimension and align the data distributions. Extensive experiments are conducted on two real CQA datasets, Toutiao and Zhihu datasets, and the results show that compared to the other advanced expert recommendation algorithms, this paper’s method achieves better results in the evaluation metrics of MAE, RMSE, Accuracy, and Recall, which fully demonstrates the effectiveness of the method in this paper to solve the data sparsity problem in expert recommendation.

In recent decades, emotion recognition has received considerable attention. As more enthusiasm has shifted to the physiological pattern, a wide range of elaborate physiological emotion data features come up and are combined with various classifying models to detect one's emotional states. To circumvent the labor of artificially designing features, we propose to acquire affective and robust representations automatically through the Stacked Denoising Autoencoder (SDA) architecture with unsupervised pre-training, followed by supervised fine-tuning. In this paper, we compare the performances of different features and models through three binary classification tasks based on the Valence-Arousal-Dominance (VAD) affection model. Decision fusion and feature fusion of electroencephalogram (EEG) and peripheral signals are performed on hand-engineered features; data-level fusion is performed on deep-learning methods. It turns out that the fusion data perform better than the two modalities. To take advantage of deep-learning algorithms, we augment the original data and feed it directly into our training model. We use two deep architectures and another generative stacked semi-supervised architecture as references for comparison to test the method's practical effects. The results reveal that our scheme slightly outperforms the other three deep feature extractors and surpasses the state-of-the-art of hand-engineered features. [ABSTRACT FROM AUTHOR]

It has been proved that long noncoding RNA (lncRNA) plays critical roles in many human diseases. Therefore, inferring associations between lncRNAs and diseases can contribute to disease diagnosis, prognosis and treatment. To overcome the limitation of traditional experimental methods such as expensive and time-consuming, several computational methods have been proposed to predict lncRNA-disease associations by fusing different biological data. However, the prediction performance of lncRNA-disease associations identification needs to be improved. In this study, we propose a computational model (named LDICDL) to identify lncRNA-disease associations based on collaborative deep learning. It uses an automatic encoder to denoise multiple lncRNA feature information and multiple disease feature information, respectively. Then, the matrix decomposition algorithm is employed to predict the potential lncRNA-disease associations. In addition, to overcome the limitation of matrix decomposition, the hybrid model is developed to predict associations between new lncRNA (or disease) and diseases (or lncRNA). The ten-fold cross validation and de novo test are applied to evaluate the performance of method. The experimental results show LDICDL outperforms than other state-of-the-art methods in prediction performance. [ABSTRACT FROM AUTHOR]

Ultra-high voltage transformers have compact structure and low insulation margin. The field strength concentration area was likely to cause "rapid development" discharge faults under the high voltage.In this paper, the high-frequency second impulse current signals under different discharge development stages were obtained by building a "rapid development" creeping discharge test platform for transformer oil-paper insulation based on the "pin-plate" defect.According to the difference of its phase-amplitude and phase-discharge maps, the discharge process was divided into three parts: "early stage"（sparse discharge), "middle stage" (large-value intense discharge) and "late stage"（small-amplitude dense discharge). On the basis of 18 characteristic parameters extracted from its PRPD map, an evaluation method of Stacked Denoising Autoencoder joint Support Vector Machine was constructed, and the development status of the "rapidly developing" creeping discharge of the transformer was evaluated, which shows a good effect. [ABSTRACT FROM AUTHOR]

In an extensive-scale surveillance system, the quality of the surveillance camera installed varies. This variation of surveillance camera produces different image quality in terms of resolution, illumination, and noise. The quality of the captured image depends on the surveillance camera hardware, placement and orientation, and the surrounding light. A pixelated, low illumination and noisy image produced by a low-quality surveillance camera causes critical issues for video surveillance face recognition systems. To address these issues, a deep learning image enhancement (DLIE) model is proposed. By utilizing a deep learning architecture such as a convolutional neural network (CNN) and a denoising autoencoder, the image quality can be enhanced. The DLIE model is able to improve image resolution and illumination and reduce noise in an image. There are two deep learning blocks (DLB) in the DLIE model, which are DLB1 and DLB2. Both DLBs are arranged in parallel so that all the stated problems can be addressed simultaneously. DLB1 is proposed to address the occurrence of pixelated images by reconstructing a low-resolution image into a high-resolution image using a CNN. DLB2 used the capability of a denoising autoencoder to reconstruct the corrupted image into a clean image by enhancing the dark and noisy images. The output of each DLB is fused using image fusion to obtain the optimum image quality. The image is evaluated using the peak to signal noise ratio (PSNR) and structural similarity index (SSIM). The enhanced image from the DLIE model exhibits superior quality compared to the original image ranging from 13.3625 to 22.7728 for PSNR and 0.6207 to 0.8155 for SSIM. [ABSTRACT FROM AUTHOR]

Aiming at the problem of low detection rate of rare attacks in traditional machine learning methods when dealing with unbalanced massive high-dimensional data, an intrusion detection model based on deep learning and random forest algorithm is proposed. In order to avoid the problems of low classification accuracy, poor stability and low detection rate of rare attacks when traditional random forests face high-dimensional data and unbalanced data, Generative Adversarial Network and Stacked Denoising Autoencoder are introduced into the Random Forest algorithm for improvement. The rare attack data set is input into the GAN neural network to generate a new attack sample to improve the uneven distribution of network intrusion data in the sample set. The deep-stacked SDAE extracts the distribution rules of the network data layer by layer, and combines the coefficient penalty and reconstruction error of each coding layer to determine the features related to the intrusion behavior in the high-dimensional data. The forest decision tree is constructed based on the characteristic data after dimension reduction. The experimental results using the UNSW-NB15 data set show that compared with SVM, KNN, CNN, LSTM, and DBN methods, the overall detection accuracy of GAN-SDAE-RF has increased by 9.39% on average, and the FPR and FNR have decreased by 9% and 15.24% on average. The detection rates on Shellcode, Backdoor, and Worms have increased by 26.8%, 27.98%, 27.85%, and 39.97% respectively. [ABSTRACT FROM AUTHOR]

Background: Plant long non-coding RNAs (lncRNAs) play vital roles in many biological processes mainly through interactions with RNA-binding protein (RBP). To understand the function of lncRNAs, a fundamental method is to identify which types of proteins interact with the lncRNAs. However, the models or rules of interactions are a major challenge when calculating and estimating the types of RBP. Results: In this study, we propose an ensemble deep learning model to predict plant lncRNA-protein interactions using stacked denoising autoencoder and convolutional neural network based on sequence and structural information, named PRPI-SC. PRPI-SC predicts interactions between lncRNAs and proteins based on the k-mer features of RNAs and proteins. Experiments proved good results on Arabidopsis thaliana and Zea mays datasets (ATH948 and ZEA22133). The accuracy rates of ATH948 and ZEA22133 datasets were 88.9% and 82.6%, respectively. PRPI-SC also performed well on some public RNA protein interaction datasets. Conclusions: PRPI-SC accurately predicts the interaction between plant lncRNA and protein, which plays a guiding role in studying the function and expression of plant lncRNA. At the same time, PRPI-SC has a strong generalization ability and good prediction effect for non-plant data. [ABSTRACT FROM AUTHOR]

High-frequency oscillations (HFOs) recorded by the intracranial electroencephalography (iEEG) are the promising biomarkers of epileptogenic zones. Accurate detection of HFOs is the key to pre-operative assessment for epilepsy. Due to the subjective bias caused by manual features and the class imbalance between HFOs and false HFOs, it is difficult to obtain satisfactory detection performance by the existing methods. To solve these problems, we put forward a novel method to accurately detect HFOs based on the stacked denoising autoencoder (SDAE) and the ensemble classifier with sample weight adjusting factors. First, the adjustable threshold of Hilbert envelopes is proposed to isolate the events of interest (EoIs) from background activities. Then, the SDAE network is utilized to automatically extract features of EoIs in the time-frequency domain. Finally, the AdaBoost-based support vector machine ensemble classifier with sample weight adjusting factors is devised to separate HFOs from EoIs by using the extracted features. These adjusting factors are used to solve the class imbalance problem by adjusting sample weights when learning the base classifiers. Our HFO detection method is evaluated by using clinical iEEG data recorded from 20 patients with medically refractory epilepsy. The experimental results show that our detection method outperforms some existing methods in terms of sensitivity and false discovery rate. In addition, the HFOs detected by our method are effective for localizing seizure onset zones. [ABSTRACT FROM AUTHOR]

The collaborative filtering method is widely used in the traditional recommendation system. The collaborative filtering method based on matrix factorization treats the user's preference for the item as a linear combination of the user and the item latent vectors, and cannot learn a deeper feature representation. In addition, the cold start and data sparsity remain major problems for collaborative filtering. To tackle these problems, some scholars have proposed to use deep neural network to extract text information, but did not consider the impact of long-distance dependent information and key information on their models. In this paper, we propose a neural collaborative filtering recommender method that integrates user and item auxiliary information. This method fully integrates user-item rating information, user assistance information and item text assistance information for feature extraction. First, Stacked Denoising Auto Encoder is used to extract user features, and Gated Recurrent Unit with auxiliary information is used to extract items' latent vectors, respectively. The attention mechanism is used to learn key information when extracting text features. Second, the latent vectors learned by deep learning techniques are used in multi-layer nonlinear networks to learn more abstract and deeper feature representations to predict user preferences. According to the verification results on the MovieLens data set, the proposed model outperforms other traditional approaches and deep learning models making it state of the art. [ABSTRACT FROM AUTHOR]

To improve the accuracy of terminal area meteorological scene pattern recognition, this study adopts a clustering model based on Stacked Denoising Autoencoder. Noise is added to the input layer, and a three-layer autoencoder is constructed for greedy layer-wise training. The reduced-dimensional features are used as inputs for clustering to achieve meteorological scene pattern recognition. The method is validated using one year of meteorological data from Tianjin Binhai International Airport. Traditional similarity distance measures such as Euclidean distance, Hamming distance, and Manhattan distance are used with both K-medoids and FCM clustering methods. The results show that the similarity measure based on SDAE performs the best in both K-medoids and FCM clustering, with a difference rate of 22.4%, 12%, 17.7%, and 24.8%, 10.7%, 11.8% compared to other similarity measures, respectively. It also has the shortest computation time, demonstrating that the SDAE-based measure and clustering achieve the best performance. Ultimately, eight meteorological scenes are identified with clear and distinct classifications. [ABSTRACT FROM AUTHOR]

Different types of outliers have existed in the monitoring data of wind turbines, which are not conducive to the follow-up data mining. However, the complex inner characteristics of the monitoring data pose major challenges to detect the outliers. To address this problem, an unsupervised outlier detection approach combining stacked denoising autoencoder (SDAE) and density-grid-based clustering method is proposed. First, the characteristics of the outliers in supervisory control and data acquisition data caused by different reasons are analyzed. Then, the SDAE is utilized to extract features by training the original data. Furthermore, the density-grid-based clustering method is applied to achieve the clustering results. Window width is added to classify the outliers as isolated outliers, missing data, and fault data according to the duration of abnormal data. The monitoring data of four wind turbines are sampled as the training data to demonstrate the effectiveness of the proposed method. The results show that the proposed model can effectively identify the isolated outliers, missing data, and fault information in the high dimensional data set by unsupervised learning.

Compressive sensing is a desirable technique to acquire and reconstruct signals at sub-Nyquist rates. Recently, several deep learning-based studies on solving the compressive sensing problem have been carried out, which dramatically reduce the intensive computational complexity of the traditional greedy or convex recovery algorithms and even improve the signal recovery performance. However, as the signal size increases, most of these methods recover signals block by block due to the large computational complexity and memory consumption, which usually imposes block effect on the recovered signals. To deal with this issue, in this paper, we apply a tensor decomposition method named Tensor-Train (TT) on the neural network, by which the number of parameters is reduced by a tremendous factor and the computational complexity is further decreased so that the large signals can be recovered as a whole. In particular, the TT-decomposition is jointly applied on a stacked denoising autoencoder (SDA) network called TT-SDA in this paper. The experiments indicate that the proposed TT-SDA network can preserve the reconstruction performance of the conventional SDA network and outperform the traditional methods, especially with low measurement rates. Meanwhile, it can also significantly reduce the computational complexity and occupied memory space, which becomes a time and memory efficient method in compressive sensing problem.

The mains signal is a complex fusion of various electrical equipment load signals in a building. In the non-intrusive load monitoring recognition, our main aim is to be able to extract as much load features as possible from the complex aggregate mains signal in a simpler way through a computer vision-based approach as opposed to the powers series signal approach. Power series methods, which are one dimensional in nature, suffer from poor aggregate and load signal feature localization necessitating a larger training dataset spanning very long time periods and normally require signal formatting and pre-processing. We use Gramian angular summation fields to transform the power series into a reduced image dataset that contains a rich set of localized signal features. A computer vision approach allows us to capture as much information as possible, and then propose an image-based mains load recognition system with high performance. In this paper for the entire recognition system, we use convolutional neural networks that very well adapted to vision recognition. The load signal image disaggregation is achieved through the powerful stacked denoising autoencoder noise extraction network. To test the proposed system, some simulations and comparisons are carried out and the results show that our easier to handle method can achieve acceptable performance. [ABSTRACT FROM AUTHOR]

Accurate fault prediction of rolling bearing can predict the operation condition in advance, which is an important means to ensure the safety and reliability of rotating machinery. Aimed at the data processing of rolling bearing vibration signal with multi-fault and long time series, an intelligent fault prediction model based on gate recurrent unit and hybrid autoencoder is proposed in this paper. Firstly, vibration signals of multi-faults are brought into multi-layer gate recurrent unit model for multi-step and multi-variable time series prediction. Secondly, variational autoencoder is used for data augmentation of fault samples. Thirdly, the augmented fault samples are brought into stacked denoising autoencoder for noise reduction and fault prediction. Finally, fault prediction results of rolling bearing can be achieved on the basis of gate recurrent unit and hybrid autoencoder of variational autoencoder and stacked denoising autoencoder. The bearing datasets of Case Western Reserve University are used to verify the effectiveness of the proposed method. Comparative experiment results show that the proposed fault prediction model has more accurate time series prediction result and higher fault prediction accuracy than other deep learning models. With 98.68% accuracy of fault prediction, the proposed fault prediction model can be taken as an effective tool for intelligent predictive maintenance of rolling bearing. [ABSTRACT FROM AUTHOR]

Assessing the operators cognitive workload status in a human - machine system based on EEG signals can effectively prevent the operator's performance from degrading. This study proposes a novel inter - subject CW classifier, E - SDAE, to adapt the variations of the EEC feature distributions across multiple subjects. The E - SDAE includes two cascade - connected modules: high level personalized feature abstractions and decision fusion. The feature filters employ the base learner SDAE to abstract the EEG features from a group of individuals. The supervised CW classifier exploites the random characteristics of ELM Lo fuse the filtered EEG abstractions under the implementation of Q - statistics. The classification rate achieves 0. 635 3 and 0. 674 7 for Task 1 and Task 2, respectively, and are significantly better than several conventional CW estimators. The time complexity calculation results show that the computational workload of the E - SDAE is also acceptable for high - dimensional EEG features. [ABSTRACT FROM AUTHOR]

In recent decades, emotion recognition has received considerable attention. As more enthusiasm has shifted to the physiological pattern, a wide range of elaborate physiological emotion data features come up and are combined with various classifying models to detect one’s emotional states. To circumvent the labor of artificially designing features, we propose to acquire affective and robust representations automatically through the Stacked Denoising Autoencoder (SDA) architecture with unsupervised pre-training, followed by supervised fine-tuning. In this paper, we compare the performances of different features and models through three binary classification tasks based on the Valence-Arousal-Dominance (VAD) affection model. Decision fusion and feature fusion of electroencephalogram (EEG) and peripheral signals are performed on hand-engineered features; data-level fusion is performed on deep-learning methods. It turns out that the fusion data perform better than the two modalities. To take advantage of deep-learning algorithms, we augment the original data and feed it directly into our training model. We use two deep architectures and another generative stacked semi-supervised architecture as references for comparison to test the method’s practical effects. The results reveal that our scheme slightly outperforms the other three deep feature extractors and surpasses the state-of-the-art of hand-engineered features.