Deep Learning Autoencoder Research Articles

ECG signal de-noising based on deep learning auto encoder and discrete wavelet transform

ECG is very important tool for diagnosis of heart disease, this signal is suffered from different types of noises such as baseline wander (BW), muscle artifact (MA) and electrode motion (EM) , which lead to wrong interpretation. In order to prevent or reduce the effect of these noises, different approaches have been applied to enhance the ECG signal. In this paper, we have proposed a new method for ECG signal de-noising based on deep learning Auto encoder (DL-DAE) and wavelet transform named as (WT-DAE). The proposed system (WT-DAE) is constructed from two stages, in the first stage, the wavelet transform is used to isolate the most significant coefficient of the signal (approximation sub-band) from de-tails coefficients (details sub-band). The details coefficients is fed to new proposed threshold method , which is used to evaluate the threshold value according to the feature of ECG signal, this threshold value is used to threshold the detail coefficients, in order to remove the details noise that is contained as high frequencly component , then invers wavelet transform is used to reconstruct the signal . Different wavelet filters and threshold functions are applied in this stage. The second stage of signal de-noising is performed by using DAE method, which is designed for reconstruct the de-noised sig-nal. The proposed DAE model is constructed from 14 layers of convolutional, relu and max_ pooling layer with different parameters. We perform training and testing the model with MIT-BIH ECG database and the performance of the pro-posed system is evaluated by terms of MSE, RMSE, PRD and PSNR. The experimental results are compared with other approaches and show that, the proposed system demonstrated the superiority for de-noising ECG signal.Â

A Deep Learning Method to Accelerate the Disaster Response Process

This paper presents an end-to-end methodology that can be used in the disaster response process. The core element of the proposed method is a deep learning process which enables a helicopter landing site analysis through the identification of soccer fields. The method trains a deep learning autoencoder with the help of volunteered geographic information and satellite images. The process is mostly automated, it was developed to be applied in a time- and resource-constrained environment and keeps the human factor in the loop in order to control the final decisions. We show that through this process the cognitive load (CL) for an expert image analyst will be reduced by 70%, while the process will successfully identify 85.6% of the potential landing sites. We conclude that the suggested methodology can be used as part of a disaster response process.

Deep Learning for Adverse Event Detection from Web Search

Adverse event detection is critical for many real-world applications including timely identification of product defects, disasters, and major socio-political incidents. In the health context, adverse drug events account for countless hospitalizations and deaths annually. Since users often begin their information seeking and reporting with online searches, examination of search query logs has emerged as an important detection channel. However, search context - including query intent and heterogeneity in user behaviors - is extremely important for extracting information from search queries, and yet the challenge of measuring and analyzing these aspects has precluded their use in prior studies. We propose DeepSAVE, a novel deep learning framework for detecting adverse events based on user search query logs. DeepSAVE uses an enriched variational autoencoder encompassing a novel query embedding and user modeling module that work in concert to address the context challenge associated with search-based detection of adverse events. Evaluation results on three large real-world event datasets show that DeepSAVE outperforms existing detection methods as well as comparison deep learning auto encoders. Ablation analysis reveals that each component of DeepSAVE significantly contributes to its overall performance. Collectively, the results demonstrate the viability of the proposed architecture for detecting adverse events from search query logs.

Employing a novel 2-gram subgroup intra pattern (2GSIP) with stacked auto encoder for membrane protein classification.

Cell membrane proteins play an essentially significant function in manipulating the behaviour of cells. Examination of amino acid sequences can put forward useful insights into the tertiary structures of proteins and their biological functions. One of the important problems in amino acid analysis is the uncertainty to establish a digital coding system to better reflect the properties of amino acids and their degeneracy. In order to overcome the demerits, the proposed method is a novel representation of protein sequences that incorporates a new feature named 2-gram subgroup intra pattern. The functional types of membrane protein classification will be supportive to explain the biological functions of membrane proteins. For classification, Stacked Auto Encoder Deep learning method is applied. The performance of the proposed method is evaluated on two benchmark data sets. The results were experimented using the Self-consistency test, Accuracy, Specificity, Sensitivity, Mathew's correlation coefficient, Jackknife test and Independent data set are the tests in which the proposed method outperformed other existing techniques generally used in literatures.

Reliability of Jack-up against Punch-through using Failure State Intelligent Recognition Technique

The preload operation of jack-up in complex multi-layered foundation requires enhanced understanding of its behaviour in punchthrough accident and suitable safety analysis tools for the assessment of their reliability for a particular site. In this study, reliability analysis model of jack-up against punch-through is established considering structural uncertainty. In order to identify the failure state, an improved reliability solution method has been developed based on Sparse Auto-Encoder (SAE) deep learning network model. Sparse self-coding algorithm is used to the training of the deep network, and Softmax regression model is established to solve the identification and classification problem of the output layer. The first application of the technique was the study of HYSY 941 jack-up platform. More specifically, numerical calculations of structure ultimate bearing capacity have been undertaken, and the influence of model parameters on the prediction accuracy of the failure state is discussed. The results show that implicit performance function can be constructed accurately using SAE-MC method by reflecting the relationship between different critical safety state and structural vulnerability. Compared with traditional BP neural network, deep learning network has higher prediction accuracy to failure probability. The dynamic risk grade in the process of preload operation can be determined quantitatively using the reliability analysis method mentioned in this paper.

A Deep Learning-based Ranking Approach for Microblog Retrieval

Today, Twitter has become one of the most popular micro-blogging service with a large amount of information on various topics produced by millions of users every day. When searching for useful information in Twitter, users need to assess high quality content that meets their needs. However, the study of effective information retrieval in such microblog is still a challenge because there is a large difference in the quality level of relevant tweets returned in the search results for a given query. Therefore, looking for an effective microblog retrieval requires distinguishing the high-quality tweet content among thousands of results. Existing works are based on hand-crafted features (e.g., number of re-tweets, number of followers, etc.) using hard-hand-engineering to indicate the quality of tweets. In this paper, we focus on the problem of ranking tweets, and particularly on retrieving high quality content. We propose a ranking approach based on k-means clustering to distinguish high quality from low quality tweets. The clustering algorithm is based on learning features from deep learning autoencoder and hand-crafted features from tweets’ content and authors’ profiles. We used information gain as a feature importance measure to find the optimal set of features having stronger power in clustering the data. By conducting a pilot feature analysis study, we demonstrate the impact of the learned features to identify tweets’ quality in the clustering process. Our experimental results show that the integration of learned features has shown significant improvement in the quality of clustering and especially on the ranking performance compared to the use of hand-crafted features only.

Model-Based: End-to-End Molecular Communication System Through Deep Reinforcement Learning Auto Encoder

Molecular communication (MC) system is an emerging technology for nanoscale networks. Therefore, there is a requirement to develop a new end-to-end MC model, which may deliver new perceptions into the aspect of these nanoscale networks. This paper aims to implement the MC framework as an end-to-end deep reinforcement learning (DRL) auto encoder (AE). The technique enables training of the MC system without any information about the actual channel (medium) model. For training the receiver and transmitter, the proposed techniques are supervised learning and DRL, respectively. The results show that the performance of the DRL autoencoder (AE) based system achieves nearly the same performance as the traditional modulation and demodulation methods in term of bit-error-rate (BER) under the Gaussian noise channel but with less complexity. The proposed technique can also be joint with the other coding methods to improve their performance.

Exploring convolutional auto-encoders for representation learning on networks

A multitude of important real-world or synthetic systems possess network structure. Extending learning techniques such as neural networks to process such non-euclidean data is therefore an important direction for machine learning research. However, till very recently this domain has received comparatively low levels of attention. There is no straight forward application of machine learning to network data as machine learning tools are designed for $i.i.d$ data, simple euclidean data or grids. To address this challenge the technical focus of this dissertation is on use of graph neural networks for Network Representation Learning (NRL) i.e. learning vector representations of nodes in networks. Learning vector embeddings of graph-structured data is similar to embedding complex data into low-dimensional geometries. After the embedding process is completed, drawbacks associated with graph structured data are overcome. The current inquiry proposes two deep learning auto-encoder based approaches for generating node embeddings. The drawbacks in existing auto-encoder approaches such as shallow architectures and excessive parameters are tackled in the proposed architectures using fully convolutional layers. Extensive experiments are performed on publicly available benchmark network data-sets to highlight the validity of this approach.

Automatic Modulation Classification Using Deep Learning Based on Sparse Autoencoders With Nonnegativity Constraints

We demonstrate a novel method for the automatic modulation classification based on a deep learning autoencoder network, trained by a nonnegativity constraint algorithm. The learning algorithm aims to constrain the negative weights, learns features that amount to a part-based representation of data, and disentangles a more meaningful hidden structure. The performance of this algorithm is tested on the fourth-order cumulants of the modulated signals. The results indicate that the autoencoder with nonnegativity constraint (ANC) improves the sparsity and minimizes the reconstruction error in comparison with the conventional sparse autoencoder. The classification accuracy of an ANC based deep network shows improved accuracy under limited signal length and fading channel.

Fuzzy risk stratification and risk assessment model for clinical monitoring in the ICU

BackgroundThe decisions that clinicians make in intensive care units (ICUs) based on monitored parameters reflecting physiological deterioration are of major medical and biomedical engineering interest. These parameters have been investigated and assessed for their usefulness in risk assessment. MethodsTotally, 127 ICU adult patients were studied. They were selected from a MIMIC II Waveform Database Matched Subset and had continuous monitoring of heart rate, invasive blood pressure, and oxygen saturation. The monitored data were dimension reduced using deep learning autoencoders and then used to train a support vector machine model (SVM). A combination of methods including fuzzy c-means clustering (FCM), and a random forest (RF) was used to determine the risk levels. ResultsWhen classifying patients into stable or deteriorating groups the main performance parameter was the receiver operating characteristics (ROC). The area under the ROC (AUROC) was 93.2 (95% CI (92.9–93.4)) with sensitivity and specificity values of 0.80 and 0.89, respectively.The suggested fuzzy risk levels using the combined method of the FCM clustering and RF achieved an accuracy of 1 (0.9999, 1), with both sensitivity and specificity values equal to 1. ConclusionsThe potential for using models in risk assessment to estimate a patient's physiological status, stable or deteriorating, within 4 h has been demonstrated. The study was based on retrospective analysis and further studies are needed to evaluate the impact on clinical outcomes using this model.

Automatic tissue characterization of air trapping in chest radiographs using deep neural networks.

Significant progress has been made in recent years for computer-aided diagnosis of abnormal pulmonary textures from computed tomography (CT) images. Similar initiatives in chest radiographs (CXR), the common modality for pulmonary diagnosis, are much less developed. CXR are fast, cost effective and low-radiation solution to diagnosis over CT. However, the subtlety of textures in CXR makes them hard to discern even by trained eye. We explore the performance of deep learning abnormal tissue characterization from CXR. Prior studies have used CT imaging to characterize air trapping in subjects with pulmonary disease; however, the use of CT in children is not recommended mainly due to concerns pertaining to radiation dosage. In this work, we present a stacked autoencoder (SAE) deep learning architecture for automated tissue characterization of air-trapping from CXR. To our best knowledge this is the first study applying deep learning framework for the specific problem on 51 CXRs, an F-score of ≈ 76.5% and a strong correlation with the expert visual scoring (R=0.93, p =<; 0.01) demonstrate the potential of the proposed method to characterization of air trapping.

Deep Learning of Part-Based Representation of Data Using Sparse Autoencoders With Nonnegativity Constraints.

We demonstrate a new deep learning autoencoder network, trained by a nonnegativity constraint algorithm (nonnegativity-constrained autoencoder), that learns features that show part-based representation of data. The learning algorithm is based on constraining negative weights. The performance of the algorithm is assessed based on decomposing data into parts and its prediction performance is tested on three standard image data sets and one text data set. The results indicate that the nonnegativity constraint forces the autoencoder to learn features that amount to a part-based representation of data, while improving sparsity and reconstruction quality in comparison with the traditional sparse autoencoder and nonnegative matrix factorization. It is also shown that this newly acquired representation improves the prediction performance of a deep neural network.

A Hybrid Malicious Code Detection Method based on Deep Learning

In this paper, we propose a hybrid malicious code detection scheme based on AutoEncoder and DBN (Deep Belief Networks). Firstly, we use the AutoEncoder deep learning method to reduce the dimensionality of data. This could convert complicated high-dimensional data into low dimensional codes with the nonlinear mapping, thereby reducing the dimensionality of data, extracting the main features of the data; then using DBN learning method to detect malicious code. DBN is composed of multilayer Restricted Boltzmann Machines (RBM, Restricted Boltzmann Machine) and a layer of BP neural network. Based on unsupervised training of every layer of RBM, we make the output vector of the last layer of RBM as the input vectors of BP neural network, then conduct supervised training to the BP neural network, finally achieve the optimal hybrid model by fine-tuning the entire network. After inputting testing samples into the hybrid model, the experimental results show that the detection accuracy getting by the hybrid detection method proposed in this paper is higher than that of single DBN. The proposed method reduces the time complexity and has better detection performance.