Imbalanced Text Sentiment Classification Based on Multi-Channel BLTCN-BLSTM Self-Attention.

With the continuous expansion of the field of natural language processing, researchers have found that there is a phenomenon of imbalanced data distribution in some practical problems, and the excellent performance of most methods is based on the assumption that the samples in the dataset are data balanced. Therefore, the imbalanced data classification problem has gradually become a problem that needs to be studied. Aiming at the sentiment information mining of an imbalanced short text review dataset, this paper proposed a fusion multi-channel BLTCN-BLSTM self-attention sentiment classification method. By building a multi-channel BLTCN-BLSTM self-attention network model, the sample after word embedding processing is used as the input of the multi-channel, and after fully extracting features, the self-attention mechanism is fused to strengthen the sentiment to further fully extract text features. At the same time, focus loss rebalancing and classifier enhancement are combined to realize text sentiment predictions. The experimental results show that the optimal F1 value is up to 0.893 on the Chnsenticorp-HPL-10,000 corpus. The comparison and ablation of experimental results, including accuracy, recall, and F1-measure, show that the proposed model can fully integrate the weight of emotional feature words. It effectively improves the sentiment classification performance of imbalanced short-text review data.

A Lightweight Hybrid Deep Learning Privacy Preserving Model for FC-Based Industrial Internet of Medical Things.

The Industrial Internet of Things (IIoT) is gaining importance as most technologies and applications are integrated with the IIoT. Moreover, it consists of several tiny sensors to sense the environment and gather the information. These devices continuously monitor, collect, exchange, analyze, and transfer the captured data to nearby devices or servers using an open channel, i.e., internet. However, such centralized system based on IIoT provides more vulnerabilities to security and privacy in IIoT networks. In order to resolve these issues, we present a blockchain-based deep-learning framework that provides two levels of security and privacy. First a blockchain scheme is designed where each participating entities are registered, verified, and thereafter validated using smart contract based enhanced Proof of Work, to achieve the target of security and privacy. Second, a deep-learning scheme with a Variational AutoEncoder (VAE) technique for privacy and Bidirectional Long Short-Term Memory (BiLSTM) for intrusion detection is designed. The experimental results are based on the IoT-Botnet and ToN-IoT datasets that are publicly available. The proposed simulations results are compared with the benchmark models and it is validated that the proposed framework outperforms the existing system.

End-to-End Model-Based Detection of Infants with Autism Spectrum Disorder Using a Pretrained Model.

In this paper, we propose an end-to-end (E2E) neural network model to detect autism spectrum disorder (ASD) from children's voices without explicitly extracting the deterministic features. In order to obtain the decisions for discriminating between the voices of children with ASD and those with typical development (TD), we combined two different feature-extraction models and a bidirectional long short-term memory (BLSTM)-based classifier to obtain the ASD/TD classification in the form of probability. We realized one of the feature extractors as the bottleneck feature from an autoencoder using the extended version of the Geneva minimalistic acoustic parameter set (eGeMAPS) input. The other feature extractor is the context vector from a pretrained wav2vec2.0-based model directly applied to the waveform input. In addition, we optimized the E2E models in two different ways: (1) fine-tuning and (2) joint optimization. To evaluate the performance of the proposed E2E models, we prepared two datasets from video recordings of ASD diagnoses collected between 2016 and 2018 at Seoul National University Bundang Hospital (SNUBH), and between 2019 and 2021 at a Living Lab. According to the experimental results, the proposed wav2vec2.0-based E2E model with joint optimization achieved significant improvements in the accuracy and unweighted average recall, from 64.74% to 71.66% and from 65.04% to 70.81%, respectively, compared with a conventional model using autoencoder-based BLSTM and the deterministic features of the eGeMAPS.

Multi-Path and Group-Loss-Based Network for Speech Emotion Recognition in Multi-Domain Datasets.

Speech emotion recognition (SER) is a natural method of recognizing individual emotions in everyday life. To distribute SER models to real-world applications, some key challenges must be overcome, such as the lack of datasets tagged with emotion labels and the weak generalization of the SER model for an unseen target domain. This study proposes a multi-path and group-loss-based network (MPGLN) for SER to support multi-domain adaptation. The proposed model includes a bidirectional long short-term memory-based temporal feature generator and a transferred feature extractor from the pre-trained VGG-like audio classification model (VGGish), and it learns simultaneously based on multiple losses according to the association of emotion labels in the discrete and dimensional models. For the evaluation of the MPGLN SER as applied to multi-cultural domain datasets, the Korean Emotional Speech Database (KESD), including KESDy18 and KESDy19, is constructed, and the English-speaking Interactive Emotional Dyadic Motion Capture database (IEMOCAP) is used. The evaluation of multi-domain adaptation and domain generalization showed 3.7% and 3.5% improvements, respectively, of the F1 score when comparing the performance of MPGLN SER with a baseline SER model that uses a temporal feature generator. We show that the MPGLN SER efficiently supports multi-domain adaptation and reinforces model generalization.

Temperature Prediction Based on Bidirectional Long Short-Term Memory and Convolutional Neural Network Combining Observed and Numerical Forecast Data.

Weather is affected by a complex interplay of factors, including topography, location, and time. For the prediction of temperature in Korea, it is necessary to use data from multiple regions. To this end, we investigate the use of deep neural-network-based temperature prediction model time-series weather data obtained from an automatic weather station and image data from a regional data assimilation and prediction system (RDAPS). To accommodate such different types of data into a single model, a bidirectional long short-term memory (BLSTM) model and a convolutional neural network (CNN) model are chosen to represent the features from the time-series observed data and the RDAPS image data. The two types of features are combined to produce temperature predictions for up to 14 days in the future. The performance of the proposed temperature prediction model is evaluated by objective measures, including the root mean squared error and mean bias error. The experiments demonstrated that the proposed model combining both the observed and RDAPS image data is better in all performance measures for all prediction periods compared with the BLSTM-based model using observed data and the CNN-BLSTM-based model using RDAPS image data alone.

Method for Fault Diagnosis of Temperature-Related MEMS Inertial Sensors by Combining Hilbert-Huang Transform and Deep Learning.

In this paper, we propose a novel method for fault diagnosis in micro-electromechanical system (MEMS) inertial sensors using a bidirectional long short-term memory (BLSTM)-based Hilbert-Huang transform (HHT) and a convolutional neural network (CNN). First, the method for fault diagnosis of inertial sensors is formulated into an HHT-based deep learning problem. Second, we present a new BLSTM-based empirical mode decomposition (EMD) method for converting one-dimensional inertial data into two-dimensional Hilbert spectra. Finally, a CNN is used to perform fault classification tasks that use time-frequency HHT spectrums as input. According to our experimental results, significantly improved performance can be achieved, on average, for the proposed BLSTM-based EMD algorithm in terms of EMD computational efficiency compared with state-of-the-art algorithms. In addition, the proposed fault diagnosis method achieves high accuracy in fault classification.

Deep-Learning-Based Detection of Infants with Autism Spectrum Disorder Using Auto-Encoder Feature Representation.

Autism spectrum disorder (ASD) is a developmental disorder with a life-span disability. While diagnostic instruments have been developed and qualified based on the accuracy of the discrimination of children with ASD from typical development (TD) children, the stability of such procedures can be disrupted by limitations pertaining to time expenses and the subjectivity of clinicians. Consequently, automated diagnostic methods have been developed for acquiring objective measures of autism, and in various fields of research, vocal characteristics have not only been reported as distinctive characteristics by clinicians, but have also shown promising performance in several studies utilizing deep learning models based on the automated discrimination of children with ASD from children with TD. However, difficulties still exist in terms of the characteristics of the data, the complexity of the analysis, and the lack of arranged data caused by the low accessibility for diagnosis and the need to secure anonymity. In order to address these issues, we introduce a pre-trained feature extraction auto-encoder model and a joint optimization scheme, which can achieve robustness for widely distributed and unrefined data using a deep-learning-based method for the detection of autism that utilizes various models. By adopting this auto-encoder-based feature extraction and joint optimization in the extended version of the Geneva minimalistic acoustic parameter set (eGeMAPS) speech feature data set, we acquire improved performance in the detection of ASD in infants compared to the raw data set.

Clarifying urban flood response characteristics and improving interpretable flood prediction with sparse data considering the coupling effect of rainfall and drainage pipeline siltation.

With climate warming and accelerated urbanisation, severe urban flooding has become a common problem worldwide. Frequent extreme rainfall events and the siltation of drainage pipes further increase the burden on urban drainage networks. However, existing studies have not fully considered the effects of rainfall and pipeline siltation on the response characteristics of flooding when constructing numerical models of urban flooding simulations. To solve this problem, a surface-subsurface coupling model was constructed by combining the Saint-Venant equation, Manning equation, a one-dimensional pipeline model (SWMM), and a two-dimensional surface overflow model (LISFLOOD-FP). Then, the SWMM model considering pipeline siltation and the two-dimensional surface overflow model (LISFLOOD-FP) were coupled with the flow exchange governing equation, and the urban flooding response characteristics considering the coupling effect of "rainfall and drainage pipeline siltation" were analysed. To enhance the solvability of waterlogging prediction, an intelligent prediction model of urban flooding based on Bayes-CNN-BLSTM was established by combining a convolutional neural network (CNN), bidirectional long short-term memory neural network (BLSTM), Bayesian optimisation (Bayes), and an interpretable loss function error correction method. The actual rainfall events and flooding processes recorded by the monitoring equipment at Huizhou University were used to calibrate and verify the model. The results show that in the Rainfall 1 and Rainfall 2 scenarios, the overload rates of the pipelines in the current siltation scenario were 60.06 % and 68.37 %, respectively, and the proportions of overflow nodes were 24.87 % and 25.89 %, respectively. When the drainage network was initially put into operation, the overload rates of the pipeline were 36.67 % and 41.16 %, and the overflow nodes accounted for 3.05 % and 4.06 %, respectively. The inundated area and volume of urban flooding increased when the combined siltation coefficient (CSC) was 0.2; therefore, two desilting schemes were determined. Under Rainfall 1, Rainfall 2, and the four rainfall recurrence periods, the Bayes-CNN-BLSTM model had clear advantages in terms of accuracy, reliability, and robustness.

Automatic classification of arrhythmias using multi-branch convolutional neural networks based on channel-based attention and bidirectional LSTM.

Cardiac arrhythmia is an abnormal rhythm of the heartbeat and can be life-threatening Electrocardiogram (ECG) is a technology that uses an electrocardiograph machine to record a graph of the changes in electrical activity produced by the heart at each cardiac cycle. ECG can generally be used to check whether the examinee has arrhythmia, ion channel disease, cardiomyopathy, electrolyte disorder and other diseases. To reduce the workload of doctors and improve the accuracy of ECG signal recognition, a novel and lightweight automatic ECG classification method based on Convolutional Neural Network (CNN) is proposed. The multi-branch network with different receptive fields is used to extract the multi-spatial deep features of heartbeats. The Channel Attention Module (CAM) and Bidirectional Long Short-Term Memory neural network (BLSTM) module are used to filter redundant ECG features. CAM and BLSTM are beneficial for distinguishing different categories of heartbeats. In the experiments, a four-fold cross-validation technique is used to improve the generalization capability of the network, and it shows good performance on the testing set. This method divides heartbeats into five categories according to the American Advancement of Medical Instrumentation (AAMI) criteria, which is validated in the MIT-BIH arrhythmia database. The sensitivity of this method to Ventricular Ectopic Beat (VEB) is 98.5% and the F1 score is 98.2%. The precision of the Supraventricular Ectopic Beat (SVEB) is 91.1%, and the corresponding F1 score is 90.8%. The proposed method has high classification performance and a lightweight feature. In a word, it has broad application prospects in clinical medicine and health testing.

EDLM: Ensemble Deep Learning Model to Detect Mutation for the Early Detection of Cholangiocarcinoma.

The most common cause of mortality and disability globally right now is cholangiocarcinoma, one of the worst forms of cancer that may affect people. When cholangiocarcinoma develops, the DNA of the bile duct cells is altered. Cholangiocarcinoma claims the lives of about 7000 individuals annually. Women pass away less often than men. Asians have the greatest fatality rate. Following Whites (20%) and Asians (22%), African Americans (45%) saw the greatest increase in cholangiocarcinoma mortality between 2021 and 2022. For instance, 60-70% of cholangiocarcinoma patients have local infiltration or distant metastases, which makes them unable to receive a curative surgical procedure. Across the board, the median survival time is less than a year. Many researchers work hard to detect cholangiocarcinoma, but this is after the appearance of symptoms, which is late detection. If cholangiocarcinoma progression is detected at an earlier stage, then it will help doctors and patients in treatment. Therefore, an ensemble deep learning model (EDLM), which consists of three deep learning algorithms-long short-term model (LSTM), gated recurrent units (GRUs), and bi-directional LSTM (BLSTM)-is developed for the early identification of cholangiocarcinoma. Several tests are presented, such as a 10-fold cross-validation test (10-FCVT), an independent set test (IST), and a self-consistency test (SCT). Several statistical techniques are used to evaluate the proposed model, such as accuracy (Acc), sensitivity (Sn), specificity (Sp), and Matthew's correlation coefficient (MCC). There are 672 mutations in 45 distinct cholangiocarcinoma genes among the 516 human samples included in the proposed study. The IST has the highest Acc at 98%, outperforming all other validation approaches.

DeepOMe: A Web Server for the Prediction of 2'-O-Me Sites Based on the Hybrid CNN and BLSTM Architecture.

2'-O-methylations (2'-O-Me or Nm) are one of the most important layers of regulatory control over gene expression. With increasing attentions focused on the characteristics, mechanisms and influences of 2'-O-Me, a revolutionary technique termed Nm-seq were established, allowing the identification of precise 2'-O-Me sites in RNA sequences with high sensitivity. However, as the costs and complexities involved with this new method, the large-scale detection and in-depth study of 2'-O-Me is still largely limited. Therefore, the development of a novel computational method to identify 2'-O-Me sites with adequate reliability is urgently needed at the current stage. To address the above issue, we proposed a hybrid deep-learning algorithm named DeepOMe that combined Convolutional Neural Networks (CNN) and Bidirectional Long Short-term Memory (BLSTM) to accurately predict 2'-O-Me sites in human transcriptome. Validating under 4-, 6-, 8-, and 10-fold cross-validation, we confirmed that our proposed model achieved a high performance (AUC close to 0.998 and AUPR close to 0.880). When testing in the independent data set, DeepOMe was substantially superior to NmSEER V2.0. To facilitate the usage of DeepOMe, a user-friendly web-server was constructed, which can be freely accessed at http://deepome.renlab.org.

AFCNNet: Automated detection of AF using chirplet transform and deep convolutional bidirectional long short term memory network with ECG signals.

Atrial fibrillation (AF) is the most common type of cardiac arrhythmia and is characterized by the heart's beating in an uncoordinated manner. In clinical studies, patients often do not have visible symptoms during AF, and hence it is harder to detect this cardiac ailment. Therefore, automated detection of AF using the electrocardiogram (ECG) signals can reduce the risk of stroke, coronary artery disease, and other cardiovascular complications. In this paper, a novel time-frequency domain deep learning-based approach is proposed to detect AF and classify terminating and non-terminating AF episodes using ECG signals. This approach involves evaluating the time-frequency representation (TFR) of ECG signals using the chirplet transform. The two-dimensional (2D) deep convolutional bidirectional long short-term memory (BLSTM) neural network model is used to detect and classify AF episodes using the time-frequency images of ECG signals. The proposed TFR based 2D deep learning approach is evaluated using the ECG signals from three public databases. Our developed approach has obtained an accuracy, sensitivity, and specificity of 99.18% (Confidence interval (CI) as [98.86, 99.49]), 99.17% (CI as [98.85 99.49]), and 99.18% (CI as [98.86 99.49]), respectively, with 10-fold cross-validation (CV) technique to detect AF automatically. The proposed approach also classified terminating and non-terminating AF episodes with an average accuracy of 75.86%. The average accuracy value obtained using the proposed approach is higher than the short-time Fourier transform (STFT), discrete-time continuous wavelet transform (DT-CWT), and Stockwell transform (ST) based time-frequency analysis methods with deep convolutional BLSTM models to detect AF. The proposed approach has better AF detection performance than the existing deep learning-based techniques using ECG signals from the MIT-BIH database.

Tumor segmentation in automated whole breast ultrasound using bidirectional LSTM neural network and attention mechanism.

Accurate breast mass segmentation of automated breast ultrasound (ABUS) is a great help to breast cancer diagnosis and treatment. However, the lack of clear boundary and significant variation in mass shapes make the automatic segmentation very challenging. In this paper, a novel automatic tumor segmentation method SC-FCN-BLSTM is proposed by incorporating bi-directional long short-term memory (BLSTM) and spatial-channel attention (SC-attention) module into fully convolutional network (FCN). In order to decrease performance degradation caused by ambiguous boundaries and varying tumor sizes, an SC-attention module is designed to integrate both finer-grained spatial information and rich semantic information. Since ABUS is three-dimensional data, utilizing inter-slice context can improve segmentation performance. A BLSTM module with SC-attention is constructed to model the correlation between slices, which employs inter-slice context to assist segmentation for false positive elimination. The proposed method is verified on our private ABUS dataset of 124 patients with 170 volumes, including 3636 2D labeled slices. The Dice similarity coefficient (DSC), Recall, Precision and Hausdorff distance (HD) of the proposed method are 0.8178, 0.8067, 0.8292 and 11.1367. Experimental results demonstrate that the proposed method offered improved segmentation results compared with existing deep learning-based methods.

Learning emotions latent representation with CVAE for text-driven expressive audiovisual speech synthesis.

Great improvement has been made in the field of expressive audiovisual Text-to-Speech synthesis (EAVTTS) thanks to deep learning techniques. However, generating realistic speech is still an open issue and researchers in this area have been focusing lately on controlling the speech variability. In this paper, we use different neural architectures to synthesize emotional speech. We study the application of unsupervised learning techniques for emotional speech modeling as well as methods for restructuring emotions representation to make it continuous and more flexible. This manipulation of the emotional representation should allow us to generate new styles of speech by mixing emotions. We first present our expressive audiovisual corpus. We validate the emotional content of this corpus with three perceptual experiments using acoustic only, visual only and audiovisual stimuli. After that, we analyze the performance of a fully connected neural network in learning characteristics specific to different emotions for the phone duration aspect and the acoustic and visual modalities. We also study the contribution of a joint and separate training of the acoustic and visual modalities in the quality of the generated synthetic speech. In the second part of this paper, we use a conditional variational auto-encoder (CVAE) architecture to learn a latent representation of emotions. We applied this method in an unsupervised manner to generate features of expressive speech. We used a probabilistic metric to compute the overlapping degree between emotions latent clusters to choose the best parameters for the CVAE. By manipulating the latent vectors, we were able to generate nuances of a given emotion and to generate new emotions that do not exist in our database. For these new emotions, we obtain a coherent articulation. We conducted four perceptual experiments to evaluate our findings.

Forecasting air pollutant concentration using a novel spatiotemporal deep learning model based on clustering, feature selection and empirical wavelet transform.

Accurate forecasting of air pollutant concentration is of great importance since it is an essential part of the early warning system. However, it still remains a challenge due to the limited information of emission source and high uncertainties of the dynamic processes. In order to improve the accuracy of air pollutant concentration forecast, this study proposes a novel hybrid model using clustering, feature selection, real-time decomposition by empirical wavelet transform, and deep learning neural network. First, all air pollutant time series are decomposed by empirical wavelet transform based on real-time decomposition, and subsets of output data are constructed by combining corresponding decomposed components. Second, each subset of output data is classified into several clusters by clustering algorithm, and then appropriate inputs are selected by feature selection method. Third, a deep learning-based predictor, which uses three dimensional convolutional neural network and bidirectional long short-term memory neural network, is applied to predict decomposition components of each cluster. Last, air pollutant concentration forecast for each monitoring station is obtained by reconstructing predicted values of all the decomposition components. PM2.5 concentration data of Beijing, China is used to validate and test our model. Results show that the proposed model outperforms other models used in this study. In our model, mean absolute percentage error for 1, 6, 10 h ahead PM2.5 concentration prediction is 4.03%, 6.87%, and 8.98%, respectively. These outcomes demonstrate that the proposed hybrid model is a powerful tool to provide highly accurate forecast for air pollutant concentration.

A Deep Neural Network for Identifying DNA N4-Methylcytosine Sites.

Motivation: N4-methylcytosine (4mC) plays an important role in host defense and transcriptional regulation. Accurate identification of 4mc sites provides a more comprehensive understanding of its biological effects. At present, the traditional machine learning algorithms are used in the research on 4mC sites prediction, but the complexity of the algorithms is relatively high, which is not suitable for the processing of large data sets, and the accuracy of prediction needs to be improved. Therefore, it is necessary to develop a new and effective method to accurately identify 4mC sites. Results: In this work, we found a large number of 4mC sites and non 4mC sites of Caenorhabditis elegans (C. elegans) from the latest MethSMRT website, which greatly expanded the dataset of C. elegans, and developed a hybrid deep neural network framework named 4mcDeep-CBI, aiming to identify 4mC sites. In order to obtain the high latitude information of the feature, we input the preliminary extracted features into the Convolutional Neural Network (CNN) and Bidirectional Long Short Term Memory network (BLSTM) to generate advanced features. Taking the advanced features as algorithm input, we have proposed an integrated algorithm to improve feature representation. Experimental results on large new dataset show that the proposed predictor is able to achieve generally better performance in identifying 4mC sites as compared to the state-of-art predictor. Notably, this is the first study of identifying 4mC sites using deep neural network. Moreover, our model runs much faster than the state-of-art predictor.

Classification of Biomedical Texts for Cardiovascular Diseases with Deep Neural Network Using a Weighted Feature Representation Method.

This study aims to improve the performance of multiclass classification of biomedical texts for cardiovascular diseases by combining two different feature representation methods, i.e., bag-of-words (BoW) and word embeddings (WE). To hybridize the two feature representations, we investigated a set of possible statistical weighting schemes to combine with each element of WE vectors, which were term frequency (TF), inverse document frequency (IDF) and class probability (CP) methods. Thus, we built a multiclass classification model using a bidirectional long short-term memory (BLSTM) with deep neural networks for all investigated operations of feature vector combinations. We used MIMIC III and the PubMed dataset for the developing language model. To evaluate the performance of our weighted feature representation approaches, we conducted a set of experiments for examining multiclass classification performance with the deep neural network model and other state-of-the-art machine learning (ML) approaches. In all experiments, we used the OHSUMED-400 dataset, which includes PubMed abstracts related with specifically one class over 23 cardiovascular disease categories. Afterwards, we presented the results obtained from experiments and provided a comparison with related research in the literature. The results of the experiment showed that our BLSTM model with the weighting techniques outperformed the baseline and other machine learning approaches in terms of validation accuracy. Finally, our model outperformed the scores of related studies in the literature. This study shows that weighted feature representation improves the performance of the multiclass classification.

Multi-information fusion neural networks for arrhythmia automatic detection.

BACKGROUND AND OBJECTIVES: . The electrocardiograms (ECGs) are widely used to diagnose a variety of arrhythmias. Generally, the abnormalities of ECG signals mainly consist of ill-shaped ECG beat morphologies and irregular intervals. The ill-shaped ECG beat morphologies represent morphological information, while the irregular intervals denote the temporal information of ECG signals. But it is difficult to utilize morphological information and temporal information simultaneously when dealing with single ECG heartbeats, because RR interval is not contained in a single short heartbeat. Therefore, to handle this problems, a novel Multi-information Fusion Convolutional Bidirectional Recurrent Neural Network (MF-CBRNN) is proposed for arrhythmia automatic detection. METHODS: . The MF-CBRNN is designed with two parallel hybrid branches that can simultaneously focus on the beat-based information in the ECG beats and the segment-based information in the adjacent segments of the beats. A single ECG beat provides the morphological information. At the same time, the adjacent segment of the ECG beat enriches the temporal information, so the two branches are designed to exploit the multiple information contained in ECGs. Furthermore, a combination of convolutional neural networks (CNNs) and a bidirectional long short memory (BLSTM) in each branch is utilized to capture the information from the two inputs. And all the features extracted from the two branches are fused for information aggregation. RESULTS: . To evaluate the performance of the proposed model, the ECG signals from MIT-BIH databases are used for intra-patient and inter-patient paradigms. The proposed model yields an accuracy of 99.56% and an F1-score of 96.40% under the intra-patient paradigm. And it obtains an overall accuracy of 96.77% and F1-score of 77.83% under the inter-patient paradigm. CONCLUSIONS: . Compared with other studies on arrhythmia detection, our method achieves a state-of-the-art performance. It indicates that the proposed model is a promising arrhythmia detection algorithm for computer-aided diagnostic systems.

Prediction of Long Non-Coding RNAs Based on Deep Learning.

With the rapid development of high-throughput sequencing technology, a large number of transcript sequences have been discovered, and how to identify long non-coding RNAs (lncRNAs) from transcripts is a challenging task. The identification and inclusion of lncRNAs not only can more clearly help us to understand life activities themselves, but can also help humans further explore and study the disease at the molecular level. At present, the detection of lncRNAs mainly includes two forms of calculation and experiment. Due to the limitations of bio sequencing technology and ineluctable errors in sequencing processes, the detection effect of these methods is  not very satisfactory. In this paper, we constructed a deep-learning model to effectively distinguish lncRNAs from mRNAs. We used k-mer embedding vectors obtained through training the GloVe algorithm as input features and set up the deep learning framework to include a bidirectional long short-term memory model (BLSTM) layer and a convolutional neural network (CNN) layer with three additional hidden layers. By testing our model, we have found that it obtained the best values of 97.9%, 96.4% and 99.0% in F1score, accuracy and auROC, respectively, which showed better classification performance than the traditional PLEK, CNCI and CPC methods for identifying lncRNAs. We hope that our model will provide effective help in distinguishing mature mRNAs from lncRNAs, and become a potential tool to help humans understand and detect the diseases associated with lncRNAs.

Urdu Nasta'liq text recognition using implicit segmentation based on multi-dimensional long short term memory neural networks.

The recognition of Arabic script and its derivatives such as Urdu, Persian, Pashto etc. is a difficult task due to complexity of this script. Particularly, Urdu text recognition is more difficult due to its Nasta'liq writing style. Nasta'liq writing style inherits complex calligraphic nature, which presents major issues to recognition of Urdu text owing to diagonality in writing, high cursiveness, context sensitivity and overlapping of characters. Therefore, the work done for recognition of Arabic script cannot be directly applied to Urdu recognition. We present Multi-dimensional Long Short Term Memory (MDLSTM) Recurrent Neural Networks with an output layer designed for sequence labeling for recognition of printed Urdu text-lines written in the Nasta'liq writing style. Experiments show that MDLSTM attained a recognition accuracy of 98% for the unconstrained Urdu Nasta'liq printed text, which significantly outperforms the state-of-the-art techniques.