Search

The development of automatic video understanding technology is highly needed due to the rise of mass video data, like surveillance videos, personal video data. Several methods have been presented previously for automatic video captioning. But, the existing methods have some problems, like more time consume during processing a huge number of frames, and also it contains over fitting problem. This is a difficult task to automate the process of video caption. So, it affects final result (Caption) accuracy. To overcome these issues, Automatic Video Captioning using Tree Hierarchical Deep Convolutional Neural Network and attention segmental recurrent neural network-bi-directional Long Short-Term Memory (ASRNN-bi-directional LSTM) is proposed in this paper. The captioning part contains two phases: Feature Encoder and Decoder. In feature encoder phase, the tree hierarchical Deep Convolutional Neural Network (Tree CNN) encodes the vector representation of video and extract three kinds of features. In decoder phase, the attention segmental recurrent neural network (ASRNN) decode vector into textual description. ASRNN-base methods struck with long-term dependency issue. To deal this issue, focuses on all generated words from the bi-directional LSTM and caption generator for extracting global context information presented by concealed state of caption generator is local and unfinished. Hence, Golden Eagle Optimization is exploited to enhance ASRNN weight parameters. The proposed method is executed in Python. The proposed technique achieves 34.89%, 29.06% and 20.78% higher accuracy, 23.65%, 22.10% and 29.68% lesser Mean Squared Error compared to the existing methods. [ABSTRACT FROM AUTHOR]

Special load customers such as electric vehicles are emerging in modern power systems. They lead to a higher penetration of special load patterns, raising difficulty for short-term load forecasting (STLF). We propose a hierarchical STLF framework to improve load forecasting accuracy. An improved adaptive K-means clustering algorithm is designed for load pattern recognition and avoiding local sub-optimal clustering centroids. We also design bi-directional long-short-term memory neural networks with an attention mechanism to filter important load information and perform load forecasting for each recognized load pattern. The numerical results on the public load dataset show that our proposed method effectively forecasts the residential load with a high accuracy. [ABSTRACT FROM AUTHOR]

Twitter has emerged as a powerful Microblogging site in recent years. It has become one of the immediately reported platforms about any crisis in today's world. During a crisis twitter is flooded with both informational and non-informational tweets. Informational tweets are the tweets which can help humanitarian response organizations or the victims to get useful information for better decision making This paper proposes a Semantic based Ontology Infused deep learning model for Informative tweet classification during natural disasters. The proposed model makes use of both text and image features present in a tweet during a classification, a Hierarchical ensemble of Vanilla and Bi-LSTM's is used along with ontologies and external knowledge from OntoCollab for text classification, a dilated convolution-based residual neural network is used for image classification, and a meaningful statistical fusion is performed with the results obtained from text and image for tweet classification. The usage of both image and text features and contextual understanding made the model distinctive and achieved exceptional results. The proposed method has proven, after experiments, to achieve evaluation metrics better than the up-to-date existing methods; evaluated on seven different datasets and compared with eight other methods with an increase in Accuracy and decrease in FNR for seven distinguished datasets which resulted an highest F1 score of 98.416% on Sri Lanka Floods dataset. [ABSTRACT FROM AUTHOR]

As extreme weather becomes more prevalent, it has serious implications for the insurance industry, real estate owners, and historic buildings. First of all, for the extreme weather, we summarize extreme weather into three parts: extreme temperature, extreme precipitation, and extreme humidity, and we process the data by combining the actual situation and applying the improved Critic method, and we demonstrate the data processing by taking the three data of Lhasa and Luxor as examples. anomalies. Subsequently, a future extreme climate prediction model for a region based on the Bi-directional LSTM method is developed. A time-weather model was fitted based on the collected data, with the highest and lowest points as the anomalies. At the same time, we simplify the formula to derive the insurance company's revenue, thus modeling the relationship between the degree of extreme weather hazard and insurance revenue, and derive that the insurance company will not be able to underwrite when the combined hazard is greater than the ratio of the insurance claim cost to the insurance payment. We applied the resulting model to Lhasa and Luxor, and plotted the combined hazard level of the actual data and the predicted data to help the insurance company visualize the combined hazard level of the region each month, and then choose whether to underwrite the policy in that region or not. [ABSTRACT FROM AUTHOR]

Satellite time series data, widely used for land cover classification, often contain missing values due to cloud contamination, which can negatively affect classification. Numerous strategies have been developed to reconstruct the missing values to produce regular time series for machine learning classifiers, among which the compositing followed by the linear interpolation is most widely used. However, the classification improvement of linear interpolation for land cover classification has not been examined. Recently developed deep learning models such as long short term memory (LSTM) and Transformer allow such examination as they can classify time series with missing values. In this study, we compared the time series composites with missing values (without linear interpolation) and the linearly interpolated time series composites (without missing values) for land cover classification. About 18 thousand Harmonized Landsat Sentinel-2 (HLS) images acquired over Amur River Basin of China (890,308 km2) in 2021 were composited to 14 16-day periods. Two time series composites were classified, i.e., (i) the 16-day composites without interpolation that have on average 15.35% 16-day periods with missing values and (ii) the linearly interpolated 16-day composites with no missing values. The classifications showed that (1) between classifications with and without linear interpolation there was < 0.2% overall accuracy differences for the bidirectional LSTM (Bi-LSTM) and < 0.5% for the Transformer both of which were smaller than model training randomness; and (2) the computation time can be saved using composites without linear interpolation. The findings suggested that it is unnecessary to use the time-consuming linear interpolation in Bi-LSTM and Transformer-based land cover classifications. The findings were confirmed by experiments for sensitivity to the number of cloud-free composites and to different classification legends using crop type classifications. It implied the linear interpolation algorithm cannot reconstruct reliable time series for land cover classifications and historical use of such method is more about mitigating the inability of traditional classifiers to handle missing values rather than improving classifications. Linear interpolation is not necessary for LSTM and Transformer with capability to handle missing values. The training datasets and developed codes in this study are made publicly available. [ABSTRACT FROM AUTHOR]

In short-term load forecasting, deep learning models with recurrent units are widely used. For Long Short-term Memory (LSTM) models, in terms of the input features at one particular time-step, there is a weight series for each input feature. Hence, an input feature with a larger value of weight can be regarded as more important to the final forecasting result. Since the weights of the features are time-invariant, the model considers the input features as equally important at each time step, thus lack of the capability of reflecting the importance fluctuation through the time span. To tackle this issue, this paper proposes a forecasting scheme, which integrates mutual information (MI) into bi-directional long short-term memory (BILSTM) network. The MI method is used to extract the importance value of input features at different time-steps, and constitute the fluctuation matrix of the importance values for input features, which is used as coefficients to correct the original input feature. The corrected features are substituted into the BILSTM network, which further improves the forecasting accuracy. Numerical experiments have been carried out based on real-world load data. The results demonstrate the superior performance of the proposed scheme compared with a series of state-of-the-art methods and thus indicate the importance and effectiveness of capturing the fluctuation of the importance value.

Automatic seizure detection using electroen-cephalogram (EEG) can significantly expedite the diagnosis of epilepsy, thereby facilitating prompt treatment and reducing the risk of future seizures and associated complications. While most existing EEG-based epilepsy detection studies employ deep learning models, they often ignore the chronological relationships between different EEG channels. To tackle this limitation, a novel automatic epilepsy detection method is proposed, which leverages path signature and Bidirectional Long Short-Term Memory (Bi-LSTM) neural network with an attention mechanism. The path signature algorithm is used to extract discriminative features for capturing the dynamic dependencies between different channels of EEG, while Bi-LSTM with attention further analyzes the inherent temporal dependencies hidden in EEG signal features. Our method is evaluated on two public EEG databases with different sizes (CHB-MIT and TUEP) and a private database from a local hospital. Two experimental settings are used, i.e., five-fold cross-validation and leave-one-out cross-validation. Experimental results show that our method achieves 99.09%, 95.60%, and 99.87% average accuracies on CHB-MIT, TUEP with 250Hz, and TUEP with 256Hz, respectively. On the private dataset, our method also achieves 99.40% average accuracy, which outperforms other methods. Furthermore, our method exhibits robustness in patients, as demonstrated by the evaluation results of cross-patient experiments.

The growing significance of online reviews in e-commerce platforms has a profound impact on consumers' purchase decisions. Unfortunately, dishonest sellers exploit this system by engaging in deceptive practices and manipulating reviews. These fake reviews are very hard to be detected by humans. Hence, the issue of detecting fake online reviews has gained significant attention due to their prevalence. Typically, the datasets available have much higher number of real reviews as compared to fake reviews, which leads to data imbalance problem. Machine learning algorithms struggle to learn from these datasets and give biased results. However, existing studies on this topic have not adequately addressed the challenges posed by imbalanced data. They have mostly used under-sampling or over-sampling to handle imbalance in datasets, which leads to potential information loss or overfitting due to replication. Hence, to handle this class imbalance problem without using sampling techniques, this paper proposes a novel SBiLM model using Siamese Neural Network with Bidirectional LSTM and Manhattan Distance. SBiLM takes pair of reviews as inputs and computes similarity between them using distance metric. The outcomes indicate that the SBiLM model exhibits better performance in terms of both accuracy and loss, when compared with traditional deep learning models on a highly imbalanced restaurant review dataset. This presents a promising strategy for tackling the issues related to data imbalance in the domain of fake review detection. [ABSTRACT FROM AUTHOR]

In the recent scenario, it is important to extract the text from various formats, including handwritten and documents. The ability to accurately detect and recognize text, is a crucial task in many fields, such as OCR systems, document analysis, and image processing. Convolutional neural networks (CNN) and bidirectional long short-term memory (BiLSTM) are used in a unique method for text detection. Additionally, CNN consist of six sequential layers by using Adam optimizer and SoftMax. The proposed approach is designed effectively to handle the variability and complexity of text in both handwritten and document formats. The proposed approach achieved a significant improvement in text detection and recognition accuracy, with 88.5 percentage while comparing with the existing techniques of CNN and LSTM which has an accuracy of 81.8 percentage. [ABSTRACT FROM AUTHOR]

Inertial Measurement Unit (IMU) sensor-based gait phase recognition is widely used in medical and biomechanics fields requiring gait data analysis. However, there are several limitations due to the low reproducibility of IMU sensor attachment and the sensor outputs relative to a fixed reference frame. The prediction algorithm may malfunction when the user changes their walking direction. In this paper, we propose a gait phase recognition method robust to user body movements based on a floating body-fixed frame (FBF) and bi-directional long short-term memory (bi-LSTM). Data from four IMU sensors attached to the shanks and feet on both legs of three subjects, collected via the FBF method, are processed through preprocessing and the sliding window label overlapping method before inputting into the bi-LSTM for training. To improve the model's recognition accuracy, we selected parameters that influence both training and test accuracy. We conducted a sensitivity analysis using a level average analysis of the Taguchi method to identify the optimal combination of parameters. The model, trained with optimal parameters, was validated on a new subject, achieving a high test accuracy of 86.43%. [ABSTRACT FROM AUTHOR]

Cloud computing is a new technology that provides services to customers anywhere, anytime, under varying conditions and managed by a third-party cloud provider. Even though cloud computing has progressed a lot, some attacks still happen. The recent anomalous and signature attacks use clever strategies such as low-rate attacks and attacking as an authenticated user. In this paper, a novel Attack Detection and Prevention (ADAPT) method is proposed to overcome this issue. The proposed system consists of three stages. An Intrusion Detection System is initially used to check whether there is an attack or not by comparing the IP address in the Blacklist IP Database. If an attack occurs, the IP address will be added to the Blacklist IP database and blocked. The second stage uses Bi-directional LSTM and Bi-directional GRU to check the anomalous and signature attack. In the third stage, classified output is sent to reinforcement learning, if any attack occurs the IP address is added to the blacklist IP database otherwise the packets are forwarded to the user. The proposed ADAPT technique achieves a higher accuracy range than existing techniques. [ABSTRACT FROM AUTHOR]

The solution of the distribution system state estimation (DSSE) relies on the presence of physical measurements in real time. Sometimes, these measurements may not reach the control center due to the defects in meter functionality, the large communication time delays, and denial-of-service (DoS) attacks on communication channels. Addressing this issue, a novel deep learning (DL) approach is proposed by using a transformer model with the integration of bi-directional long short-term memory (Bi-LSTM) layer. The proposed model can be leveraged to forecast the unavailable measurement data, which maintains the observability of the system to obtain an accurate solution from DSSE. The superiority of the proposed model is probed by comparing it with other forecasting-based DL models at various percentages of missing measurement data. Further, the effectiveness of the proposed model is evaluated for forecasting the power injections of residential, commercial, and industrial loads as well as renewable energy sources. Finally, the solution of the DSSE is tested with forecasted data and compared with the results of the DSSE for the original measurement data. The simulations are performed on modified IEEE 13-node and IEEE 37-node test systems, and their corresponding results highlighted the superiority of the proposed model in the forecasting of the incomplete data. [ABSTRACT FROM AUTHOR]

Special load customers such as electric vehicles are emerging in modern power systems. They lead to a higher penetration of special load patterns, raising difficulty for short-term load forecasting (STLF). We propose a hierarchical STLF framework to improve load forecasting accuracy. An improved adaptive K-means clustering algorithm is designed for load pattern recognition and avoiding local sub-optimal clustering centroids. We also design bi-directional long-short-term memory neural networks with an attention mechanism to filter important load information and perform load forecasting for each recognized load pattern. The numerical results on the public load dataset show that our proposed method effectively forecasts the residential load with a high accuracy.

This study focuses on successful Forex trading by emphasizing the importance of identifying market trends and utilizing trend analysis for informed decision-making. The authors collected low-correlated currency pair datasets to mitigate multicollinearity risk. Authors developed a two-stage predictive model that combines regression and classification tasks, using the predicted closing price to determine entry and exit points. The model incorporates Bi-directional long short-term memory (Bi-LSTM) for improved price forecasting and higher highs and lower lows (HHs-HLs and LHs-LLs) to identify trend changes. They proposed an enhanced DeepSense network (DSN) with all member-based optimization (AMBO-DSN) to optimize decision variables of DSN. The performance of the models was compared to various machine learning, deep learning, and statistical approaches including support vector regressor (SVR), artificial neural network (ANN), auto-regressive integrated moving average (ARIMA), vanilla-LSTM (V-LSTM), and recurrent neural network (RNN). The optimized form of DSN using genetic algorithm (GA), particle swarm optimization (PSO), and differential evolution (DE) was compared with AMBO-DSN, yielding satisfactory results that demonstrated comparable quality to the observed trends on the original currency pairs. The effectiveness and reliability of the AMBO-DSN approach in forecasting trends for USD/EUR, AUD/JPY, and CHF/INR currency pairs were validated through statistical analysis while considering computational cost. [ABSTRACT FROM AUTHOR]

This study proposes a telemanipulation framework with two wearable IMU sensors without human skeletal kinematics. First, the states (intensity and direction) of spatial hand-guiding gestures are separately estimated through the proposed state estimator, and the states are also combined with the gesture's mode (linear, angular, and via) obtained with the bi-directional LSTM-based mode classifier. The spatial pose of the 6-DOF manipulator's end-effector (EEF) can be controlled by combining the spatial linear and angular motions based on integrating the gesture's mode and state. To validate the significance of the proposed method, the teleoperation of the EEF to the designated target poses was conducted in the motion-capture space. As a result, it was confirmed that the mode could be classified with 84.5% accuracy in real time, even during the operator's dynamic movement; the direction could be estimated with an error of less than 1 degree; and the intensity could be successfully estimated with the gesture speed estimator and finely tuned with the scaling factor. Finally, it was confirmed that a subject could place the EEF within the average range of 83 mm and 2.56 degrees in the target pose with only less than ten consecutive hand-guiding gestures and visual inspection in the first trial. [ABSTRACT FROM AUTHOR]

Sentiment analysis, commonly called opinion mining or emotion artificial intelligence (AI), employs biometrics, computational linguistics, natural language processing, and text analysis to systematically identify, extract, measure, and investigate affective states and subjective data. Sentiment analysis algorithms include emotion lexicon, traditional machine learning, and deep learning. In the text sentiment analysis algorithm based on a neural network, multi-layer Bi-directional long short-term memory (LSTM) is widely used, but the parameter amount of this model is too huge. Hence, this paper proposes a Bi-directional LSTM with a trapezoidal structure model. The design of the trapezoidal structure is derived from classic neural networks, such as LeNet-5 and AlexNet. These classic models have trapezoidal-like structures, and these structures have achieved success in the field of deep learning. There are two benefits to using the Bi-directional LSTM with a trapezoidal structure. One is that compared with the single-layer configuration, using the of the multi-layer structure can better extract the high-dimensional features of the text. Another is that using the trapezoidal structure can reduce the model's parameters. This paper introduces the Bi-directional LSTM with a trapezoidal structure model in detail and uses Stanford sentiment treebank 2 (STS-2) for experiments. It can be seen from the experimental results that the trapezoidal structure model and the normal structure model have similar performances. However, the trapezoidal structure model parameters are 35.75% less than the normal structure model. [ABSTRACT FROM AUTHOR]

An emotion is a conscious logical response that varies for different situations in women's life. These mental responses are caused by physiological, cognitive, and behavioral changes. Gender-based violence undermines the participation of women in decision-making, resulting in a decline in their quality of life. More accurate and automatic classification of women's emotions can enhance human-computer interfaces and security in real time. There are some wearable technologies and mobile applications that claim to ensure the safety of women. However, they rely on limited social action and are ineffective at ensuring women's safety when and where it is needed. In this work, a novel CDB-LSTM network has been proposed to accurately classify the emotions of women in seven different classes. The electroencephalogram (EEG) offers non-radioactive methods of identifying emotions. Initially, the EEG signals are preprocessed and they are converted into images via Time-Frequency Representation (TPR). A smoothed pseudo-Wigner-Ville distribution (SPWVD) is employed to convert the EEG time-domain signals into input images. Consequently, these converted images are given as input to the Convolutional Deep Belief Network (CDBN) for extracting the most relevant features. Finally, Bi-directional LSTM is used for classifying the emotions of women into seven classes namely: happy, relax, sad, fear, anxiety, anger, and stress. The proposed CDB-LSTM network preserves the high accuracy range of 97.27% in the validation phase. The proposed CDB-LSTM network improves the overall accuracy by 6.20% 32.98% 6.85% and 3.30% better than CNN-LSTM, Multi-domain feature fusion model, GCNN-LSTM and CNN with SVM and DT respectively. [ABSTRACT FROM AUTHOR]

Short-term passenger flow prediction of urban rail transit based on historical data mining can guide the work of station organization, train scheduling, and passenger flow induction effectively and dynamically. We propose the RFBiLSTM prediction model, which combines the random forest algorithm (RF) and bi-directional long short-term memory neural network (BiLSTM). Firstly, the time series features of the flow are obtained by clustering algorithm and correlation analysis. Secondly, RF is used to obtain the importance of the features. Finally, we compare the prediction performance of 7 models and investigate the impact of feature selection on deep learning models. Through case analysis, the prediction accuracy of RF is the highest when using a single model for prediction, and the MAPE is 0.102. When using the combined prediction model, the MAPE of RF-BiLSTM is 0.074. The accuracy of RF-BiLSTM is better compared with the results of a single model. The performance of BiLSTM is improved by 46.6% after the features are optimized using feature selection methods. The findings demonstrate the suitability of the combined prediction model RF-BiLSTM for predicting shortterm inbound passenger flow. [ABSTRACT FROM AUTHOR]

Inertial Measurement Unit (IMU) sensor-based gait phase recognition is widely used in medical and biomechanics fields requiring gait data analysis. However, there are several limitations due to the low reproducibility of IMU sensor attachment and the sensor outputs relative to a fixed reference frame. The prediction algorithm may malfunction when the user changes their walking direction. In this paper, we propose a gait phase recognition method robust to user body movements based on a floating body-fixed frame (FBF) and bi-directional long short-term memory (bi-LSTM). Data from four IMU sensors attached to the shanks and feet on both legs of three subjects, collected via the FBF method, are processed through preprocessing and the sliding window label overlapping method before inputting into the bi-LSTM for training. To improve the model’s recognition accuracy, we selected parameters that influence both training and test accuracy. We conducted a sensitivity analysis using a level average analysis of the Taguchi method to identify the optimal combination of parameters. The model, trained with optimal parameters, was validated on a new subject, achieving a high test accuracy of 86.43%.

Biomedical named entity recognition is a popular research topic in the Biosciences domain as number of biomedical articles getting published are increasing rapidly. Generic models using machine learning and deep learning techniques have been proposed for extracting these entities in the past, however there is no clear verdict on which techniques are better and how these generic models perform in a domain-specific big data scenario. In this paper, we evaluate three baseline models using the most complex BioNLP 2013 cancer genetics dataset addressing the cancer domain. A classifier ensemble, bidirectional long short-term memory (Bi-LSTM) model and the bidirectional encoder representations from transformers (BERT) model are implemented. We propose NeRBERT, a domain-specific, graphical processing unit (GPU) pre-trained language model using extra biomedical corpora extending BERTBASE. Experimental results prove the efficacy of NeRBERT as it outperforms the other three models with an F1-score gain of 12.18 pp, 8.59 pp and 5.43 pp over the ensemble, Bi-LSTM and BERT models respectively. GPUs reduce the model training time to less than half. Comparing it to existing state-of-the-art models, it performs 1.57 pp higher than the next best existing model compared, emerging as a robust biomedical and cancer phenotyping NER tagger. [ABSTRACT FROM AUTHOR]

One of the high-risk behaviors leading to severe traffic injuries is not wearing a seat belt. It is therefore very important to be able to automatically detect seat belts from surveillance images, encourage drivers to wear seat belts, and enhance passenger safety. In this paper, a novel deep neural network, Gated Bi-directional Long Short-Term Memory network with part-to-whole attention (GBL-PA), is proposed for seat belt detection from surveillance images. The innovation of our model lies in its unique diagonal sampling strategy, which meticulously captures the seat belt's fine details, typically oriented from top right to bottom left across the torso of vehicle occupants. Our framework's novelty is further encapsulated by the part-to-whole attention mechanism, which intelligently harmonizes the detailed local information from seat belt-specific patches with the broader contextual insights from the regional proposals. The pioneering design of a Gated Bi-directional LSTM network facilitates the dynamic integration of interactions across patches to deliver an optimized final prediction. The superiority of GBL-PA is established through rigorous comparison with the state-of-the-art methods on a new, large benchmark dataset comprising 14,936 images from traffic surveillance footage. Our framework demonstrates a notable improvement, achieving a mean Average Precision (mAP) of 72.3%, which surpasses the second best, YOLOX, by 0.9% mAP. This significant and consistent outperformance across various metrics underscores the transformative potential of GBL-PA in the realm of traffic safety enforcement. The source code of our framework is available at ANONYMISED. • Novel seat belt detection framework exploiting inherent properties of seat belts. • Patch Diagonal Sampling obtains patches to capture fine-grained seat belt details. • Part-to-whole Attention exploits attentive patch information under global context. • Gated Bi-LSTM captures interactions among patches to aggregate all information. • The framework greatly outperforms SoTA methods on the large, constructed dataset. [ABSTRACT FROM AUTHOR]

The outbreak of COVID-19, a little more than 2 years ago, drastically affected all segments of society throughout the world. While at one end, the microbiologists, virologists, and medical practitioners were trying to find the cure for the infection; the Governments were laying emphasis on precautionary measures like lockdowns to lower the spread of the virus. This pandemic is perhaps also the first one of its kind in history that has research articles in all possible areas as like: medicine, sociology, psychology, supply chain management, mathematical modeling, etc. A lot of work is still continuing in this area, which is very important also for better preparedness if such a situation arises in future. The objective of the present study is to build a research support tool that will help the researchers swiftly identify the relevant literature on a specific field or topic regarding COVID-19 through a hierarchical classification system. The three main tasks done during this study are data preparation, data annotation and text data classification through bi-directional long short-term memory (bi-LSTM).

The accuracy of the wearable inertia-measurement-unit (IMU)-sensor-based gesture recognition may be significantly affected by undesired changes in the body-fixed frame and the sensor-fixed frame according to the change in the subject and the sensor attachment. In this study, we proposed a novel wearable IMU-sensor-based hand-guiding gesture recognition method robust to significant changes in the subject's body alignment based on the floating body-fixed frame method and the bi-directional long short-term memory (bi-LSTM). Through comparative experimental studies with the other two methods, it was confirmed that aligning the sensor-fixed frame with the reference frame of the human body and updating the reference frame according to the change in the subject's body-heading direction helped improve the generalization performance of the gesture recognition model. As a result, the proposed floating body-fixed frame method showed a 91.7% test accuracy, confirming that it was appropriate for gesture recognition under significant changes in the subject's body alignment during gestures. [ABSTRACT FROM AUTHOR]

For the prospectus of profitability and technological advancement, solar irradiance forecasting for grid-connected solar plants is essential nowadays. However, the intermittent and stochastic nature of solar photovoltaic (PV) output has become a threat to the power security and reliability of solar-connected grids. So, in search for a stable solution, this paper proposes a hybrid model to estimate a day ahead solar irradiance by employing the full wavelet packet decomposition (FWPD) and the Bidirectional long short-term memory (BiLSTM). The FWPD extracts various frequency features and statistical characteristics of the data through decomposition process. The isolated BiLSTM network with a dropout layer is then assigned to each decomposed frequency component (sub-series), where it acts as a core predictor and obtains the futuristic value of each subseries. Finally, the final forecasting (monthly and seasonal) is obtained using the FWPD reconstruction by using averaging ensemble technique of each predicted subseries. The efficiency of the proposed model is demonstrated by comparing statistical parameters: mean absolute error (MAE), mean absolute percentage error (MAPE), root-mean-square error (RMSE), coefficient of determination (R2) and forecast skills (FS), to different contrast models: naïve (baseline) predictor, long short-term memory (LSTM), gated recurrent unit (GRU), BiLSTM and conventional wavelet transform (WT)-based BiLSTM (WTBiLSTM). The percentage improvement of proposed model in RMSE and MAPE is also discussed in this paper. In order to discuss the sensitivity with respect to difference in forecasted and observed values, various tests are conducted such as index of agreement (IA), direction change in forecasting (DC) and Diebold-Mariano hypothesis (DMH). [ABSTRACT FROM AUTHOR]

Genes are composed of DNA and each gene has a specific sequence. Recombination or replication within the gene base ends in a permanent change in the nucleotide collection in a DNA called mutation and some mutations can lead to cancer. Breast adenocarcinoma starts in secretary cells. Breast adenocarcinoma is the most common of all cancers that occur in women. According to a survey within the United States of America, there are more than 282,000 breast adenocarcinoma patients registered each 12 months, and most of them are women. Recognition of cancer in its early stages saves many lives. A proposed framework is developed for the early detection of breast adenocarcinoma using an ensemble learning technique with multiple deep learning algorithms, specifically: Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and Bi-directional LSTM. There are 99 types of driver genes involved in breast adenocarcinoma. This study uses a dataset of 4127 samples including men and women taken from more than 12 cohorts of cancer detection institutes. The dataset encompasses a total of 6170 mutations that occur in 99 genes. On these gene sequences, different algorithms are applied for feature extraction. Three types of testing techniques including independent set testing, self-consistency testing, and a 10-fold cross-validation test is applied to validate and test the learning approaches. Subsequently, multiple deep learning approaches such as LSTM, GRU, and bi-directional LSTM algorithms are applied. Several evaluation metrics are enumerated for the validation of results including accuracy, sensitivity, specificity, Mathew's correlation coefficient, area under the curve, training loss, precision, recall, F1 score, and Cohen's kappa while the values obtained are 99.57, 99.50, 99.63, 0.99, 1.0, 0.2027, 99.57, 99.57, 99.57, and 99.14 respectively. [ABSTRACT FROM AUTHOR]

This study proposes a telemanipulation framework with two wearable IMU sensors without human skeletal kinematics. First, the states (intensity and direction) of spatial hand-guiding gestures are separately estimated through the proposed state estimator, and the states are also combined with the gesture’s mode (linear, angular, and via) obtained with the bi-directional LSTM-based mode classifier. The spatial pose of the 6-DOF manipulator’s end-effector (EEF) can be controlled by combining the spatial linear and angular motions based on integrating the gesture’s mode and state. To validate the significance of the proposed method, the teleoperation of the EEF to the designated target poses was conducted in the motion-capture space. As a result, it was confirmed that the mode could be classified with 84.5% accuracy in real time, even during the operator’s dynamic movement; the direction could be estimated with an error of less than 1 degree; and the intensity could be successfully estimated with the gesture speed estimator and finely tuned with the scaling factor. Finally, it was confirmed that a subject could place the EEF within the average range of 83 mm and 2.56 degrees in the target pose with only less than ten consecutive hand-guiding gestures and visual inspection in the first trial.

Depth video sequence-based deep models for recognizing human actions are scarce compared to RGB and skeleton video sequences-based models. This scarcity limits the research advancements based on depth data, as training deep models with small-scale data is challenging. In this work, we propose a sequence classification deep model using depth video data for scenarios when the video data are limited. Unlike summarizing the frame contents of each frame into a single class, our method can directly classify a depth video, i.e., a sequence of depth frames. Firstly, the proposed system transforms an input depth video into three sequences of multi-view temporal motion frames. Together with the three temporal motion sequences, the input depth frame sequence offers a four-stream representation of the input depth action video. Next, the DenseNet121 architecture is employed along with ImageNet pre-trained weights to extract the discriminating frame-level action features of depth and temporal motion frames. The extracted four sets of feature vectors about frames of four streams are fed into four bi-directional (BLSTM) networks. The temporal features are further analyzed through multi-head self-attention (MHSA) to capture multi-view sequence correlations. Finally, the concatenated genre of their outputs is processed through dense layers to classify the input depth video. The experimental results on two small-scale benchmark depth datasets, MSRAction3D and DHA, demonstrate that the proposed framework is efficacious even for insufficient training samples and superior to the existing depth data-based action recognition methods. [ABSTRACT FROM AUTHOR]

Energy storage technologies are being used excessively in industrial applications and in automobiles. Battery state of charge (SOC) is an important metric to be monitored in these applications to ensure proper and safe functionality. Since SOC cannot be measured directly, this paper puts forth a novel machine learning architecture to improve on the existing methods of SOC estimation. This method consists of using combined stacked bi-directional LSTM and encoder–decoder bi-directional long short-term memory architecture. This architecture henceforth represented as SED is implemented to overcome the nonparallel functionality observed in traditional RNN algorithms. Estimations were made utilizing different open-source datasets such as urban dynamometer driving schedule (UDDS), highway fuel efficiency test (HWFET), LA92 and US06. The least Mean Absolute Error observed was 0.62% at 25 °C for the HWFET condition, which confirms the good functionality of the proposed architecture. [ABSTRACT FROM AUTHOR]

Bearings as the key component of most rotating machinery, responsible for major breakdowns. Therefore, this paper addresses intelligent prognostics involving remaining useful life estimation. The proposed framework is based on a deep learning model to learn the bearing degradation from vibration responses. A comprehensive feature selection strategy involving PSO (particle swarm optimization) optimization technique and feature transformations is carried out. The sensitive prognostic features set are then input to BiLSTM (bi-directional long short-term memory) network to learn long-term time dependencies in two directions. Furthermore, the attention mechanism is integrated with BiLSTM enables selective processing of information. The experimental validation is carried out by acquiring data from a high-speed rotor supported on the bearings. The results achieved higher prediction accuracy. Also, the generalization on IEEE PHM data achieves higher RUL (remaining useful life) prediction accuracy as compared to state-of-art methods. Hence, the results proved the high performance and feasibility of the proposed RUL prognostic method. [ABSTRACT FROM AUTHOR]

Deep learning (DL) based localization and Simultaneous Localization and Mapping (SLAM) has recently gained considerable attention demonstrating remarkable results. Instead of constructing hand-crafted algorithms through geometric theories, DL based solutions provide a data-driven solution to the problem. Taking advantage of large amounts of training data and computing capacity, these approaches are increasingly developing into a new field that offers accurate and robust localization systems. In this work, the problem of global localization for unmanned aerial vehicles (UAVs) is analyzed by proposing a sequential, end-to-end, and multimodal deep neural network based monocular visual-inertial localization framework. More specifically, the proposed neural network architecture is three-fold; a visual feature extractor convNet network, a small IMU integrator bi-directional long short-term memory (LSTM), and a global pose regressor bi-directional LSTM network for pose estimation. In addition, by fusing the traditional IMU filtering methods instead of LSTM with the convNet, a more time-efficient deep pose estimation framework is presented. It is worth pointing out that the focus in this study is to evaluate the precision and efficiency of visual-inertial (VI) based localization approaches concerning indoor scenarios. The proposed deep global localization is compared with the various state-of-the-art algorithms on indoor UAV datasets, simulation environments and real-world drone experiments in terms of accuracy and time-efficiency. In addition, the comparison of IMU-LSTM and IMU-Filter based pose estimators is also provided by a detailed analysis. Experimental results show that the proposed filter-based approach combined with a DL approach has promising performance in terms of accuracy and time efficiency in indoor localization of UAVs.

The prognostics and health management (PHM) plays the main role to handle the risk of failure before its occurrence. Next, it has a broad spectrum of applications including utility networks, energy storage systems (ESS), etc. However, an accurate capacity estimation of batteries in ESS is mandatory for their safe operations and decision making policy. ESS comprises of different storage mechanisms such as batteries, capacitors, etc. Consequently, the measurement of different charging profiles (CPs) has a strong relation to battery capacity. These profiles include temperature (T), voltage (V), and current (I) where the CPs patterns vary as the battery ages with cycles. Consequently, estimating a battery capacity, the conventional methods practice single channel charging profile (SCCP) and hop multiple channel CPs (MCCPs) that cause incorrect battery health estimation. To tackle these issues, this article proposes MCCPs based battery management system (BMS) to estimate batteries health/capacity through the deep learning (DL) concept where the patterns in these CPs are changed as the battery ages with time and cycles. Thus, we deeply investigate both machine learning (ML) and DL based methods to provide a concrete comparative analysis of our method. The adaptive boosting (AB) and support vector regression (SVR) are widely compared with long short-term memory (LSTM), multi-layer perceptron (MLP), bi-directional LSTM (BiLSTM), and convolutional neural network (CNN) to attain the appropriate approach for battery capacity and state of health (SOH) estimation. These approaches have a high learning capability of inter-relation between the battery capacity and variation in CPs patterns. To validate and verify the proposed technique, we use NASA battery dataset and experimentally prove that BiLSTM outperforms all the approaches and obtains the smallest error values for MAE, MSE, RMSE, and MAPE using MCCPs compared to SCCP.

Globally, human falls are the second leading cause of deaths induced due to unintentional injuries. These fatalities, in most cases, arise due to a lack of timely medication. Therefore, over the years, there has been an immense demand for systems that can quickly send fall-related information to the caretakers so that the medical relief team can reach on time. The traditional schemes for fall detection using wearable sensors such as accelerometers, gyroscopes, etc., are highly intrusive and generate high false positives in real-world conditions. Consequently, the current research directions in this domain have been toward harnessing the availability of low-cost vision sensors and the power of deep learning. To this end, in this work, we present a dual-stream fused neural network (DSFNN) for fall detection in multi-camera and 360 ∘ video streams. The DSFNN model learns to extract spatial-temporal information using two neural networks, trained independently on the RGB video sequences of fall and non-fall activities and their corresponding single dynamic images. Once trained, the model fuses the prediction scores of the two neural networks using a weighted fusion scheme to obtain the final decision. We assessed the performance of the proposed DSFNN on two multi-camera fall datasets, namely UP-Fall and URFD, and on a new in-house 360 ∘ video dataset of fall and non-fall activities. The evaluation results in terms of different performance metrics demonstrated the superiority of the proposed fall detection scheme. The framework achieved superior performance and outperformed the previous state-of-the-art fall detection methods. For further research and analysis in the fall detection domain, we will make the source code and the in-house fall dataset available to the research community on request. [ABSTRACT FROM AUTHOR]

Excessive Power Consumption (PC) and demand for power is increasing on a daily basis, due to advancements in technology, the rise in electricity-dependent machinery, and the growth of the human population. It has become necessary to predict PC in order to improve power management and co-operation between the energy used in a building and the power grid. State-of-the-art Energy Consumption Prediction (ECP) methods are limited in terms of predicting the energy effectively, due to various challenges such as weather conditions and the dynamic behaviour of occupants. Thus, to overcome the drawbacks of these methods, we present an intelligent hybrid technique that combines a Convolutional Neural Network (CNN) with a Multi-layer Bi-directional Long-short Term Memory (M-BDLSTM) method using three steps. When applied to short-term power ECP, this approach helps to provide efficient power management i.e. it can assist the supplier to produce the optimum amount of power. The first step in our proposed method integrates the pre-processing and data organisation mechanisms to refine the data and remove abnormalities. The second step employs a deep learning network, where the sequence of refined data is fed into the CNN via the M-BDLSTM network to learn the sequence pattern effectively. The third step generates the ECP/PC by comparing actual and predicted data series and evaluates the prediction using error metrics. The proposed method achieves better prediction results than existing techniques, thus demonstrating its effectiveness. Furthermore, it achieved the smallest value of the Mean Square Error (MSE) and Root Mean Square Error (RMSE) for individual household dataset using 10-fold Cross Validation (CV) and a hold-out (CV) method.

The accuracy of the wearable inertia-measurement-unit (IMU)-sensor-based gesture recognition may be significantly affected by undesired changes in the body-fixed frame and the sensor-fixed frame according to the change in the subject and the sensor attachment. In this study, we proposed a novel wearable IMU-sensor-based hand-guiding gesture recognition method robust to significant changes in the subject’s body alignment based on the floating body-fixed frame method and the bi-directional long short-term memory (bi-LSTM). Through comparative experimental studies with the other two methods, it was confirmed that aligning the sensor-fixed frame with the reference frame of the human body and updating the reference frame according to the change in the subject’s body-heading direction helped improve the generalization performance of the gesture recognition model. As a result, the proposed floating body-fixed frame method showed a 91.7% test accuracy, confirming that it was appropriate for gesture recognition under significant changes in the subject’s body alignment during gestures.

Recently, person re-identification has become one of the research hotspots in the field of computer vision and has received extensive attention in the academic community. Inspired by the part-based research of image ReID, this paper presents a novel feature learning and extraction framework for video-based person re-identification, namely, the extended global-local representation learning network (E-GLRN). Given a video sequence of a pedestrian, the holistic and local features are simultaneously extracted using the E-GLRN network. Specifically, for the global feature learning, we adopt the channel attention convolutional neural network (CNN) and the bidirectional long short-term memory (Bi-LSTM) networks, which are responsible for introducing a CNN-LSTM module to learn the features of consecutive frames. The local feature learning module relies on the key local information extraction, which is based on the Bi-LSTM networks. In order to obtain the local feature more effectively, our work defines a concept of “the main image group” by selecting three representative frames. The local feature representation of a video is obtained by exploiting the spatial contextual and appearance information of this group. The local and global features extracted in this paper are complementary and further combined into a discriminative and robust feature representation of the video sequence. Extensive experiments are conducted on three video-based ReID datasets, including the iLIDS-VID, PRID2011 and MARS datasets. The experimental results demonstrate that the proposed method outperforms state-of-the-art video-based re-identification approaches.

For the robotic positioning and navigation, visual odometry (VO) system is widely used. However, the errors of the traditional VO accumulate when the robot moves. Besides, this paper proposes a new framework to solve the problem of monocular VO, called MagicVO. Based on the convolutional neural network (CNN) and the bi-directional LSTM (Bi-LSTM), MagicVO outputs a 6-DoF absolute-scale pose at each position of the camera with a sequence of continuous monocular images as input. It does not only utilize the outstanding performance of CNN in extracting the rich features of image frames fully but also learns the geometric relationship from image sequences pre and post through Bi-LSTM to get a more accurate prediction. A pipeline of the MagicVO is shown in this paper. The MagicVO is an end-to-end system, and the results of the experiments on the KITTI and ETH datasets show that MagicVO has a better performance than the traditional VO systems in the accuracy of pose and the generalization ability.

Sign language recognition aims to recognize meaningful movements of hand gestures and is a significant solution in intelligent communication between the deaf community and hearing societies. However, until now, the current dynamic sign language recognition methods still have some drawbacks with difficulties of recognizing complex hand gestures, low recognition accuracy for most dynamic sign language recognition, and potential problems in larger video sequence data training. In order to solve these issues, this paper presents a multimodal dynamic sign language recognition method based on a deep 3-dimensional residual ConvNet and bi-directional LSTM networks, which is named as BLSTM-3D residual network (B3D ResNet). This method consists of three main parts. First, the hand object is localized in the video frames in order to reduce the time and space complexity of network calculation. Then, the B3D ResNet automatically extracts the spatiotemporal features from the video sequences and establishes an intermediate score corresponding to each action in the video sequence after feature analysis. Finally, by classifying the video sequences, the dynamic sign language is accurately identified. The experiment is conducted on test datasets, including DEVISIGN_D dataset and SLR_Dataset. The results show that the proposed method can obtain state-of-the-art recognition accuracy (89.8% on the DEVISIGN_D dataset and 86.9% on SLR_Dataset). In addition, the B3D ResNet can effectively recognize complex hand gestures through larger video sequence data, and obtain high recognition accuracy for 500 vocabularies from Chinese hand sign language.

Depth video sequence-based deep models for recognizing human actions are scarce compared to RGB and skeleton video sequences-based models. This scarcity limits the research advancements based on depth data, as training deep models with small-scale data is challenging. In this work, we propose a sequence classification deep model using depth video data for scenarios when the video data are limited. Unlike summarizing the frame contents of each frame into a single class, our method can directly classify a depth video, i.e., a sequence of depth frames. Firstly, the proposed system transforms an input depth video into three sequences of multi-view temporal motion frames. Together with the three temporal motion sequences, the input depth frame sequence offers a four-stream representation of the input depth action video. Next, the DenseNet121 architecture is employed along with ImageNet pre-trained weights to extract the discriminating frame-level action features of depth and temporal motion frames. The extracted four sets of feature vectors about frames of four streams are fed into four bi-directional (BLSTM) networks. The temporal features are further analyzed through multi-head self-attention (MHSA) to capture multi-view sequence correlations. Finally, the concatenated genre of their outputs is processed through dense layers to classify the input depth video. The experimental results on two small-scale benchmark depth datasets, MSRAction3D and DHA, demonstrate that the proposed framework is efficacious even for insufficient training samples and superior to the existing depth data-based action recognition methods.

Energy storage technologies are being used excessively in industrial applications and in automobiles. Battery state of charge (SOC) is an important metric to be monitored in these applications to ensure proper and safe functionality. Since SOC cannot be measured directly, this paper puts forth a novel machine learning architecture to improve on the existing methods of SOC estimation. This method consists of using combined stacked bi-directional LSTM and encoder–decoder bi-directional long short-term memory architecture. This architecture henceforth represented as SED is implemented to overcome the nonparallel functionality observed in traditional RNN algorithms. Estimations were made utilizing different open-source datasets such as urban dynamometer driving schedule (UDDS), highway fuel efficiency test (HWFET), LA92 and US06. The least Mean Absolute Error observed was 0.62% at 25 °C for the HWFET condition, which confirms the good functionality of the proposed architecture.

The arrival of nCoV virus took the existence of human race in the toss, thus to slow down the spread of the virus Indian Government imposed a nationwide lockdown. In this paper, we aim to find the liaison between the impacts of lockdown on slowing down the spread of the virus along with the unprecedented blow to the small cap companies enlisted in the stock market. The impact of lockdown model has been assessed by proposing a differential equation based mathematical modelling (SIQR Modelling). Along with that, the stock market effect has been analyzed using a proposed Bi-Directional LSTM model optimized using Genetic Algorithm. In this present research, we have optimized the time-stamp window using genetic algorithm, before passing through the LSTM architecture. The novelty of the paper lies in the bi-fold approach to assess the chaos of the stock market, under 2 conditional scenarios i.e. with and without the lockdown. [ABSTRACT FROM AUTHOR]

Retrieving relevant source code from large repositories is a significant and ongoing challenge in the field of software engineering, primarily due to the vast and ever-expanding amount of available code. Existing deep learning methods, although effective to some extent, exhibit limitations in capturing the intricate and complex structural information embedded within source code, which hinders their ability to provide highly accurate retrieval results. This study endeavors to tackle this prominent issue by introducing a novel and innovative approach known as the Joint Bi-directional LSTM and Graph Neural Networks (JBLG) model for source code retrieval. The central aim is to harness the combined strengths and capabilities of Bi-directional Long Short-Term Memory (LSTM) networks and Graph Neural Networks (GNNs) to significantly enhance the model's capacity to capture and interpret the complex structural characteristics intrinsic to source code. The proposed JBLG model employs a unique fusion of Bi-directional LSTM, which excels in capturing sequential and temporal dependencies within code, and GNN, which is adept at modeling the intricate graph structure of the code. By leveraging this hybrid architecture, the model aims to provide a comprehensive and highly effective solution for source code retrieval tasks. To assess the efficacy of the JBLG model, extensive experiments are conducted, and the model's performance is evaluated against well-established benchmarks, including LSTM, GNN, and ChatGPT, using two diverse datasets: CodeSearchNet and CosBench datasets. These evaluations span multiple programming languages, ensuring a comprehensive and robust assessment of the model's capabilities. The experimental results indicate that the JBLG model consistently outperforms its counterparts, including Bi-LSTM, GNN, ChatGPT, and DGMS, across various evaluation metrics. the JBLG model showcases an exceptional ability to handle and extract the intricate structural information inherent in source code, resulting in significantly enhanced retrieval accuracy. The JBLG model emerges as a highly promising solution for real-world source code retrieval applications, with the potential to revolutionize the field. The success of this model underscores the importance of combining deep learning techniques like Bi-directional LSTM and GNNs for tackling complex software engineering challenges. Furthermore, future research directions could involve exploring advanced techniques such as attention mechanisms and extending the model's applicability to other software engineering tasks like code summarization and code completion. The findings of this study are expected to have a lasting impact on the advancement of source code retrieval methodologies. [ABSTRACT FROM AUTHOR]

Media plays a vital role in the public dissemination of information about events. The rapid development of the Internet allows a quick spread of information through social networks or websites. Without the concern about the credibility of the information, the unverified or fake news is spread in social networks and reach thousands of users. Fake news is typically generated for commercial and political interests to mislead and attract readers. The spread of fake news has raised a big challenge to society. Automatic credibility analysis of news articles is a current research interest. Deep learning models are widely used for linguistic modeling. Typical deep learning models such as Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) can detect complex patterns in textual data. Long Short-Term Memory (LSTM) is a tree-structured recurrent neural network used to analyze variable-length sequential data. Bi-directional LSTM allows looking at particular sequence both from front-to-back as well as from back-to-front. The paper presents a fake news detection model based on Bi-directional LSTM-recurrent neural network. Two publicly available unstructured news articles datasets are used to assess the performance of the model. The result shows the superiority in terms of accuracy of Bi-directional LSTM model over other methods namely CNN, vanilla RNN and unidirectional LSTM for fake news detection. [ABSTRACT FROM AUTHOR]

• This paper describes approaches of caption generation that help to integrate a form of attention with variants. There are two types of variants: a soft attention technique and a hard attention technique. • Due to recent advancements in caption generation and employment of attention in machine translation, we further investigate those models, which involve salient part of a video and will generate its caption. • By visualizing what and where the attention should be focused, we interpret the results of this framework. • We also validate the advantages of attention in caption generation by using state of art performance on benchmark dataset. The study of Natural Language Generation (NLG), especially how human beings narrate the world, assists in understanding of the visual world for surveillance. Our research proposes an effective technique to axiomatically develop multi-line textual description of visual data by exploiting deep Convolution Neural Networks (CNN). Textual description of visual data aids in providing textual tags for visual information. A human can retrieve elected videos from a repository based on visual tags. Videos contain more complex and detailed information than images and provide more language data. The proposed feats-rich model encodes the visual contents to visual and facial features using CNN architecture. Encoded features are passed to two layer LSTM units with attention mechanism, reducing the number of parameters by encompassing relevant details. Experimental results on Trecvid 2016 and UET-Surveillance dataset depict that model outperforms state-of-the-art methods by scoring BLEU score of 0.35 and 0.52, respectively. [ABSTRACT FROM AUTHOR]

Recently, tremendous advances have been observed in information retrieval systems designed to search for relevant knowledge in scientific publications. Although these techniques are quite powerful, there is still room for improvement in the area of searching for metadata relating to algorithms in full-text publication datasets—for instance, efficiency-related metrics such as precision, recall, f-measure and accuracy, and other useful metadata such as the datasets deployed and the algorithmic run-time complexity. In this study, we proposed a novel deep learning-based feature engineering approach that improves search capabilities by mining algorithmic-specific metadata from full-text scientific publications. Typically, traditional term frequency-inverse document frequency (TF-IDF)-based approaches function like a 'bag of words' model and thus fail to capture either the text's semantics or the word sequence. In this work, we designed a semantically enriched synopsis of each full-text document by adding algorithmic-specific deep metadata text lines to enhance the search mechanism of algorithm search systems. These text lines are classified by our deployed deep learning-based bi-directional long short term memory (LSTM) model. The designed bi-directional LSTM model outperformed the support vector machine by 9.46%, with a 0.81 f1-score on a dataset of 37,000 algorithm-specific deep metadata text lines that had been tagged by four human experts. Lastly, we present a case study on 21,940 full-text publications downloaded from ACL (https://aclweb.org/) to show the effectiveness of deep learning-based advanced feature engineering search compared to the conventional TF-IDF-based (Lucene) search. [ABSTRACT FROM AUTHOR]

Human activity recognition (HAR) remains a challenging yet crucial problem to address in computer vision. HAR is primarily intended to be used with other technologies, such as the Internet of Things, to assist in healthcare and eldercare. With the development of deep learning, automatic high-level feature extraction has become a possibility and has been used to optimize HAR performance. Furthermore, deep-learning techniques have been applied in various fields for sensor-based HAR. This study introduces a new methodology using convolution neural networks (CNN) with varying kernel dimensions along with bi-directional long short-term memory (BiLSTM) to capture features at various resolutions. The novelty of this research lies in the effective selection of the optimal video representation and in the effective extraction of spatial and temporal features from sensor data using traditional CNN and BiLSTM. Wireless sensor data mining (WISDM) and UCI datasets are used for this proposed methodology in which data are collected through diverse methods, including accelerometers, sensors, and gyroscopes. The results indicate that the proposed scheme is efficient in improving HAR. It was thus found that unlike other available methods, the proposed method improved accuracy, attaining a higher score in the WISDM dataset compared to the UCI dataset (98.53% vs. 97.05%).

The drive for smarter, greener, and more livable cities has led to research towards more effective solar energy forecasting techniques and their integration into traditional power systems. However, the availability of real-time data, data storage, and monitoring has become challenging. This research investigates a method based on Bi-directional LSTM (BDLSTM) neural network. BDLSTM takes into account the data's past and future context. The future hidden layer takes input in ascending order while the past hidden layer evaluates the input in decreasing order, making BDLSTM relevant in analyzing the input data's past context and evaluating future predictions. The eleven-year (2010–2020) weather dataset used for this paper was acquired from NASA. Two pre-processing approaches, Automatic Time Series Decomposition (ATSD) and Pearson correlation, were used to remove the noisy values from the residual components and for feature selection, respectively. To ensure storage and reuse of data, the architecture includes a cloud-based server for data management and reuse for future predictions. Popular in multi-energy systems, the cloud-based server also serves as a platform for monitoring predicted solar energy data. The metrics values and results obtained have demonstrated that the BDLSTM performs efficiently on the available data. Data from two separate climatic horizons proved the study's quality and reliability. • A deep LSTM network, covering BDLSTM, LSTM and MLP is proposed for hourly forecasting of solar energy. • Automatic time-series decomposition is designed to smooth meteorological variables, addressing the seasonality and trend problems so as to improve the overall accuracy of the model. • To ensure the management of multi-energy systems, cloud-based architecture is established to store and process collected data for energy monitoring, control, and future prediction. [ABSTRACT FROM AUTHOR]

Short text categorization is a crucial issue to many applications, e.g., Information Retrieval, Question-Answering System, MRI Database Construction and so forth. Many researches focus on data sparsity and ambiguity issues in short text categorization. To tackle these issues, we propose a novel short text categorization strategy based on abundant representation, which utilizes Bi-directional Recurrent Neural Network(Bi-RNN) with Long Short-Term Memory(LSTM) and topic model to catch more contextual and semantic information. Bi-RNN enriches contextual information, and topic model discovers more latent semantic information for abundant text representation of short text. Experimental results demonstrate that the proposed model is comparable to state-of-the-art neural network models and method proposed is effective. [ABSTRACT FROM AUTHOR]