This work addresses the intricate task of predicting the prices of diverse financial assets, including stocks, indices, and cryptocurrencies, each exhibiting distinct characteristics and behaviors under varied market conditions. To tackle the challenge effectively, a novel hybridized architecture, AE-GRU, integrating the encoder-decoder principle with GRU is designed. The experimentation involves multiple activation functions and hyperparameter tuning. With extensive experimentation and enhancements applied to AE-LSTM, the proposed AE-GRU architecture still demonstrates significant superiority in forecasting the annual prices of volatile financial assets from multiple sectors mentioned above. Thus, the novel AE-GRU architecture emerges as a superior choice for price prediction across diverse sectors and fluctuating market scenarios by extracting important non-linear features of financial data and retaining the long-term context from past observations.

Few-shot object detection (FSOD) aims to address the challenge of requiring a substantial amount of annotations for training in conventional object detection, which is very labor-intensive. However, existing few-shot methods achieve high precision with the sacrifice of time-consuming for exhaustive fine-tuning, or take poor performance in novel-class adaptation. We presume the major reason is that the valuable correlation feature among different categories is insufficiently exploited, hindering the generalization of knowledge from base to novel categories for object detection. In this paper, we propose Few-Shot object detection via Correlation-RPN and Transformer Encoder-Decoder (CRTED), a novel training network to learn object-relevant features of inter-class correlation and intra-class compactness while suppressing object-agnostic features in the background with limited annotated samples. And we also introduce a 4-way tuple-contrast training strategy to positively activate the training progress of our object detector. Experiments over two few-shot benchmarks (Pascal VOC, MS-COCO) demonstrate that, our proposed CRTED without further fine-tuning can achieve comparable performance with current state-of-the-art fine-tuned works. The codes and pre-trained models will be released.

Image captioning aims at generating meaningful verbal descriptions of a digital image. Our paper focuses on the classic encoder-decoder deep learning model that consists of several components – sub-networks, each performing a separate task that, combined, form an effective caption generator. We investigate image feature extractors, recurrent neural networks, word embedding models, and word generation layers and discuss how each component influences the captioning model’s overall performance. Our experiments are performed on the MS COCO 2014 dataset. The results help design efficient models with optimal combinations of their components.

Cardiovascular disease is the leading cause of death worldwide. A key factor in assessing the risk of cardiovascular disease is left ventricular functional evaluation. Left ventricular (LV) systolic function is evaluated by measuring the left ventricular ejection fraction (LVEF) using echocardiography data. Therefore, quick and accurate left ventricle segmentation is important for estimating the LVEF. However, it is difficult to accurately segment the left ventricle due to changes in the shape and area of the left ventricle during cardiac cycles. In this study, we proposed a framework that considers changes in the shape and area of the left ventricle during the cardiac cycle by applying the convolutional long short-term memory (CLSTM) approach. In addition, we evaluated the left ventricular segmentation and multidimensional quantification of the proposed system in comparison to manual and automated segmentation methods. In addition, to assess the validity of CLSTM, the values of multi-dimensional quantification metrics were compared and analyzed using graphs and Bland–Altman plots on a frame-by-frame basis. We demonstrated that the CLSTM method effectively segments the left ventricle by considering the LV activity. In conclusion, we demonstrated that LV segmentation based on our framework may be utilized to accurately estimate LVEF values.

With the development of artificial intelligence, the ability of capturing the background characteristics of hyperspectral imagery (HSI) is improved and promising performance in hyperspectral anomaly detection (HAD) tasks is yielded. However, existing methods proposed in recent years still suffer from certain limitations: 1) constraints are lacking in the deep feature learning process in terms of the issue for absence of prior background and anomaly information. 2) hyperspectral anomaly detectors with traditional self-supervised deep learning methods fail to ensure prioritized reconstruction of the background. 3) architectures of fully connected deep network in hyperspectral anomaly detectors lead to low utilization of spatial information and destruction for original spatial relationship of hyperspectral imagery, and disregard spectral correlation between adjacent pixels. 4) hypotheses or assumptions for background and anomaly distributions restrict the performance of many hyperspectral anomaly detectors because the distributions of background land covers are usually complex and not assumable in real-world hyperspectral imagery. With the consideration of above problems, in this paper, we propose a novel fully convolutional auto-encoder based on dual clustering and latent feature adversarial consistency (FCAE-DCAC) for HAD which is carried out in a self-supervised learning-based processing. Firstly, the density-based spatial clustering of applications with noise algorithm and connected component analysis are utilized for successive spectral and spatial clustering to obtain more precise prior background and anomaly information, which facilitates the separation between background and anomaly samples during the training of our method. Subsequently, a novel fully convolutional auto-encoder (FCAE) integrated with spatial-spectral joint attention (SSJA) is proposed to enhance the utilization of spatial information and augment feature expression. In addition, a latent feature adversarial consistency network is proposed to achieve pure background reconstruction with the ability of learning actual background distribution in hyperspectral imagery. Finally, a triplet loss is introduced to enhance the separability between background and anomaly, and the reconstruction residual serves as the anomaly detection result. We evaluate the proposed method on seven groups of real-world hyperspectral datasets, and the experimental results confirmed the effectiveness and superior performance of the proposed method versus nine state-of-the-art methods.

The accurate forecasting of solar power output is essential for ensuring the stable and efficient operation of grid-connected photovoltaic (PV) systems, especially when faced with variable weather patterns. This research focuses on the integration of Machine Learning (ML) methodologies and climatic parameters to predict solar panel energy generation, with a specific emphasis on addressing consumption-production imbalances. Leveraging a dataset sourced from the Kaggle platform, the study is conducted in the context of Estonia, aiming to optimize solar energy utilization in this geographic region. The dataset, obtained from Kaggle, encompasses comprehensive information on climatic variables, including sunlight intensity, temperature, and humidity, alongside corresponding solar panel energy output. Leveraging machine learning algorithms, including XGBoost regression and neural networks, our model aims to identify complex patterns and relationships within the datasets. By tailoring the model to Estonia's climatic nuances, we seek to enhance the accuracy of energy production forecasts and, consequently, better manage the challenges associated with consumption-production imbalances. Furthermore, the research investigates the adaptability of the proposed model to diverse climatic conditions, ensuring its applicability for similar endeavors in other geographical locations. The performance of the forecasting model will be evaluated using error metrics like mean absolute error (MAE) and root mean squared error (RMSE). This study leverages a comprehensive dataset from Kaggle and utilizes advanced machine learning techniques to generate valuable insights. These insights can inform the development of sustainable energy policies and practices, ultimately leading to a more efficient and reliable renewable energy infrastructure.

Responsible, efficient, and environmentally conscious energy consumption practices are increasingly essential for ensuring the reliability of the modern electricity grid. This study focuses on leveraging time series analysis to improve forecasting accuracy, crucial for various application domains where real-world time series data often exhibit complex, non-linear patterns. Our approach advocates for utilizing Long Short Term Memory (LSTM) and Bidirectional LSTM (BiLSTM) models for precise time series forecasting. To ensure a fair evaluation, we compare the performance of our proposed approach with traditional neural networks, time-series forecasting methods, and conventional decline curves. Additionally, individual models based on LSTM, Bi-LSTM, and other machine learning methods are implemented for a comprehensive assessment. Experimental results consistently demonstrate that our proposed model outperforms all benchmarking methods in terms of mean absolute error (MAE) across most datasets. Addressing the imbalance between activations by consumer and prosumer groups, our predictions show superior performance compared to several traditional forecasting methods, such as the Autoregressive Integrated Moving Average model (ARIMA) and Seasonal Autoregressive Integrated Moving Average model (SARIMA). Specifically, the root mean square error (RMSE) of Bi-LSTM is 5.35%, 46.08%, and 50.6% lower than LSTM, ARIMA, and SARIMA, respectively, on the May test data.

Understanding the distribution of rock glaciers provides key information for investigating and recognizing the status and changes of the periglacial environment. Deep learning algorithms and red-green-blue (RGB) bands from high-resolution satellite images have been extensively employed to map rock glaciers. However, the near-infrared (NIR) band offers rich spectral information and sharp edge features that could significantly contribute to semantic segmentation tasks, but it is rarely utilized in constructing rock glacier identification models due to the limitation of three input bands for classical semantic segmentation networks, like DeeplabV3+. In this study, a dual-encoder DeeplabV3+ network (DEDNet) was designed to overcome the flaws of classical DeeplabV3+ network (CDNet) when identifying rock glaciers using multispectral remote sensing images by extracting spatial and spectral features from RGB and NIR bands, respectively. This network, trained with manually labeled rock glacier samples from QiLian Mountains, established a model with accuracy, precision, recall, specificity, and mIoU (Mean Intersection over Union) of 0.9131, 0.9130, 0.9270, 0.9195, and 0.8601, respectively. The well-trained model was applied to identify new rock glaciers in a test region, achieving a producer's accuracy of 93.68% and a user's accuracy of 94.18%. Furthermore, the model was employed to two study areas in northern Tien Shan (Kazakhstan) and Daxue Shan (Hengduan Shan, China) with high accuracy, which proved that the DEDNet offers an innovative solution to more accurately map rock glaciers on a larger scale due to its robustness across diverse geographic regions.

Pepper leaf segmentation plays a crucial role in monitoring pepper leaf diseases in various backgrounds and ensuring the healthy growth of peppers. However, existing transformer-based segmentation methods suffer from computational inefficiency, excessive parameterization, and limited utilization of edge information. To tackle these challenges, we propose an adaptive multi-scale MLP framework, named AMS-MLP, which combines the multi-path aggregation module (MPAM) and the multi-scale context relation mask module (MCRD) to refine the object boundaries in pepper leaf segmentation. AMS-MLP consists of an encoder-based network, an adaptive multi-scale MLP (AM-MLP) module, and a decoder network. In the encoder network, the MPAM module effectively fuses five-scale features to generate a single-channel mask, improving the accuracy of pepper leaf boundary extraction. The AM-MLP module divides the input features into two branches: the global multi-scale MLP branch captures long-range dependencies between image information, while the local multi-scale MLP branch focuses on extracting local feature maps. Adaptive attention mechanism is designed to dynamically adjust the weights of global and local features. The decoder network incorporates the MCRD module into the convolutional layer, enhancing the extraction of boundary features. To verify the performance of the proposed method, we conducted extensive experiments on three pepper leaf datasets with different backgrounds. The results demonstrate mIoU scores of 97.39%, 96.91%, and 97.91%, as well as F1 scores of 98.29%, 97.86%, and 98.51%, respectively. Comparative analysis with U-Net and state-of-the-art models reveals that the proposed method dramatically improves the accuracy and efficiency of pepper leaf image segmentation.

Quantum computing (QC) has opened the door to advancements in machine learning (ML) tasks that are currently implemented in the classical domain. Convolutional neural networks (CNNs) are classical ML architectures that exploit data locality and possess a simpler structure than fully-connected multi-layer perceptrons (MLPs) without compromising the accuracy of classification. However, the concept of preserving data locality is usually overlooked in the existing quantum counterparts of CNNs, particularly for extracting multi-features in multidimensional data. In this paper, we present a multidimensional quantum convolutional classifier (MQCC) that performs multidimensional and multi-feature quantum convolution with average and Euclidean pooling and thus adapting the CNN structure to a variational quantum algorithm (VQA). Average pooling is based on the quantum Haar transform (QHT) and Euclidean pooling is based on partial quantum measurement. The experimental work was conducted using multidimensional data to validate the correctness and demonstrate the scalability of the proposed method utilizing both noisy and noise-free quantum simulations. We evaluated the MQCC model with reference to reported work on state-of-the-art quantum simulators from IBM Quantum and Xanadu using a variety of standard ML datasets. The experimental results showed favorable characteristics of our proposed techniques compared to existing work with respect to a number of quantitative metrics such as the number of training parameters, cross-entropy loss, classification accuracy, circuit depth, and quantum gate count.

Iterative Stochastic decoding is an alternative to standard fixed-point decoding of Low-density parity check codes (LDPC) to reduce inter-node routing. In this paper, we propose a Flexible Field-Programmable Gate Array (FPGA)-Based Stochastic decoding (SD) hardware architecture for LDPC codes in the Fifth-Generation (5G) New Radio (NR) Standard that supports decoding of set of various code rates. This decoder’s runtime flexibility is desirable to switch a better performing code rate automatically based on the channel conditions without the extra time needed for reprogramming of FPGA. An offline design method is implemented to generate the hardware description language (HDL) code description of the decoder for the required code-rate set, which is further synthesized and integrated into the Xilinx Kintex-7 series FPGA board to determine the hardware resource utilisation and processing throughput. The Synopsys design tools were employed during both the simulation and synthesis stages, in combination with TSMC 65-nm CMOS standard cell technology, to facilitate comparative analysis. Compared with state-of-the-art designs, the proposed architecture reduces hardware utilization by up to 26% and achieved energy efficiency by 52%.

Health management in industrial systems is more critical in maintenance management and it plays an important role in productivity, fault diagnosis, safety, efficiency, and economy in manufacturing industries. Early detection of faults in machinery may increase the effectiveness of maintenance actions and will avoid unwanted consequences in process operations and maintenance. Existing fault diagnosis methods have limitations such as insufficient accuracy, slow detection rate, and handling large and complex data sets. In this digital age, Industry 4.0 techniques have been applied across all fields, including the condition monitoring of machines. This research addresses the gaps in traditional fault diagnosis by using deep learning, a modern AI technique effective for diagnosing faults in various machines. Deep learning algorithms Multilayer Perceptron (MLP), Coevolutionary Neural Network (CNN), and Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) are tested for fault diagnosis using vibration datasets collected from Spectra Quest Machinery Fault Simulator (SQMFS). In this research work, NI-DAQ (National Instruments- Data Acquisition) system, accelerometer, and LabVIEW software are used to collect vibration signals. Preprocessing of the signals has been done using a sampling strategy, shuffling, standardization, and reshaping data augmentation. The result shows that MLP accuracy in the prediction of fault is 0.9, CNN reached 0.95, and RNN and LSTM with 0.57 and 0.45 respectively. The high performance of CNN is due to its ability to effectively capture spatial patterns in vibration data which is crucial for fault diagnosis in rotating machinery followed by MLP due to its faster convergence during training. However, when scaling the data, MLP outperformed CNN, demonstrating superior adaptability to increased data complexity and volume. Due to the need for larger datasets and temporal patterns in the vibration data, which RNN and LSTM are designed to handle, they resulted in a lower accuracy. This study shows that CNN has given better results than other deep learning algorithms MLP, RNN, and LSTM in fault diagnosis of rotating machinery. Future research could explore the application of these techniques to different types of machinery and fault conditions.

To enhance the safety of power grid operations, this study proposes a high-precision short-term photovoltaic power prediction method that integrates information from surrounding pho-tovoltaic stations and the Conv-LSTM-ATT model. In the deep learning prediction model, not only is numerical weather prediction (NWP) data from the target photovoltaic station used as input features, but also highly correlated features from nearby photovoltaic stations are incor-porated. The research begins by analyzing the correlation between irradiance and power se-quences, along with distance factors, to calculate a composite similarity index between the target and other regional photovoltaic stations. Stations with high similarity indices are then selected as data sources. Subsequently, Bayesian optimization techniques are employed to find the optimal data fusion ratios. Ultimately, using the selected data, power prediction mod-eling is conducted via the Conv-LSTM-ATT deep neural network. Experimental results con-firm the superiority of the proposed model, which demonstrates higher predictive accuracy compared to three other classical models. The data fusion strategy determined by Bayesian optimization significantly enhances prediction accuracy, reducing the root mean square error (RMSE) of the test set by 20.04%, 28.24%, and 30.94% for three weather types, respectively.

Rice has become the second major food among rural people. It is one of the most nutrient and low-cost food available in Asia. Rice is composed of two major subspecies as Japonica and Indica and it belongs to the Poaceae family. The World Bank predicts that rice demand will grow 51% faster than population growth in 2025. It is high time to develop a model which predicts the type of infected paddy leaf and avoids the loss of grains. In this paper, the proposed hybrid CNN model, three CNN networks with different layer properties are trained parallel. The extracted feature from these CNN networks is combined to provide the classification of the type of infected leaf. Fieldwork was carried and about 340 samples are collected from five different types of classes. The proposed hybrid model was tested with both the fieldwork dataset and also with an available dataset on the internet. The experimental results show that the proposed model outperforms the other CNN network and yields an accuracy value of 98%, a precision value of 95%, and a misclassification rate of 4% respectively.

This paper proposed emotion recognition methods for consecutive facial images, and implements the inference of neural network model on a field-programmable gate array (FPGA). The proposed emotion recognition methods are based on a neural network model architecture that combines convolution neural networks (CNNs), long short-term memory (LSTM), and fully connected neural networks (FCNNs), called CLDNN or ConvLSTM-FCN. This type of neural network mod-el can analyze the local feature sequences obtained through convolution of data, making it suita-ble for processing time-series data such as consecutive facial images. In this paper, sequences of facial images are sampled from videos corresponding to various emotional state of subjects. The sampled images are then pre-processed with the processes includes facial detection, greyscale conversion, resize, and data augmentation if necessary. The 2-D CNN in ConvLSTM-FCN is used for feature extraction for these pre-processed facial images. These sequences of facial im-age features are time sequences with dependent properties between the elements within them. The LSTM is then used to model these time sequences followed by fully connected neural net-works (FCNNs) for classification. The proposed consecutive facial emotion recognition method achieves an average recognition rate of 99.51% on RAVDESS dataset, 87.80% on BAUM-1s dataset and 96.82% on eNTERFACE’05 data set, using 10-fold cross-validation on the PC. Some com-parisons of recognition accuracies between the proposed method and the other existing related works are conducted in this paper. According to the comparisons, the proposed emotion recog-nition methods outperform the existing related researches. The proposed emotion recognition methods are then implemented on an FPGA chip using the neural network model inference algo-rithms in this paper, and the accuracies of the experiments conducted on the FPGA chip are iden-tical to those obtained on the PC. This verifies that the proposed neural network model imple-mented on the FPGA chip performs well.

The Dst index is the geomagnetic storm index used to measure the energy level of geomagnetic storms, and the prediction of this index is of great significance for the geomagnetic storm study and the solar activity. In contrast to traditional numerical modelling techniques, machine learning, which has emerged in decades ago based on rapidly developing computer hardware and software and artificial intelligence methods, has been unprecedentedly developed in geophysics, especially solar-terrestrial space physics. This study chooses two machine learning models, the LSTM (Long-Short Time Memory, LSTM) and EMD-LSTM model (Empirical Mode Decomposition, EMD), to model and predict the Dst index. By building the Dst index data series from 2018-2023, two models were built to fit and predict the data. Firstly, we evaluated the influences of the learning rate and the amount of training data on the prediction accuracy of the LSTM model, and finally, 10-3 was thought as the optimal learning rate; secondly, the two models were used to predict the Dst index in the solar active and quiet periods, respectively, and the RMSE (Root Mean Square Error) of the LSTM model in the active period is 7.34 nT, the CC (correlation coefficient) is 0.96, those of the quiet period are 2.64nT and 0.97; the RMSE and r of EMD-LSTM model are 8.87nT and 0.93 in active time and 3.29nT and 0.95 in the quiet time. Finally, the prediction accuracy of the LSTM model in short time period is slightly better than the EMD-LSTM model. However, there will be a problem of prediction lag, which the EMD-LSTM model can then solve, and can better predict the geomagnetic storm.

This paper introduces an innovative approach to lipreading, leveraging a web application designed to generate subtitles for videos where the speaker's mouth is visible and a comprehensive literature review that precedes the discussion, encompassing a thorough examination of various lipreading methods employed over the past decade. Our method employs a powerful deep learning model, featuring a 3D-convolution network and bidirectional LSTM, enabling accurate sentence-level predictions based solely on visual lip movements. With an impressive accuracy of 97%, our model is trained using pre-segmented lips regions, transformed into animated GIFs for effective pre-training. This work stands as a significant contribution to the evolving landscape of lipreading research, offering a practical and accurate solution for real-world applications.

Glaucoma, a global leading cause of blindness, necessitates accurate segmentation of the optic disc (OD) and optic cup (OC) for effective screening. However, existing segmentation methods based on convolutional neural networks (CNNs) suffer from high computational complexity and long inference times. In this paper, we propose an end-to-end segmentation method for the OD and OC using a lightweight MobileNetv3 as the feature extraction network. Our approach incorporates boundary branches and an adversarial learning network to achieve multi-label segmentation of the OD and OC. We evaluate the proposed method on three publicly available datasets: Drishti-GS, RIM-ONE-r3, and REFUGE. The experimental results demonstrate segmentation accuracies of 0.974/0.900, 0.966/0.875, and 0.962/0.880 for the OD and OC on the respective datasets, while significantly reducing inference time.

In the application domain, accurate word sense identification is crucial for improving the performance of machine translation, information retrieval and end-to-end communication tasks. However, word polysemy is a major obstacle to accurate semantic identification. Therefore, word semantic disambiguation has always been one of the key challenges in natural language processing and has attracted the attention of a large number of researchers. This research proposes an innovative disambiguation algorithm based on the large-scale Bert model and the Polly encoder framework, and introduces WordNet as a benchmark for word semantic. By exploiting the ability of the pre-trained model to extract and learn semantic information, and using a specially designed forward propagation algorithm and loss function to fine-tune the large-scale Bert model, the model has high Accuracy and robustness. In this research, several experiments were conducted on the Semcor 3.0 semantic dataset. The experimental results show that the model proposed in this research shows excellent performance on the Semcor test set, with an Accuracy of 86.1% and an F1 score of 0.847, which is a significant improvement over the traditional model.

Artificial neural networks are a powerful tool for managing data that is difficult to process and interpret. This paper presents the study of artificial neural networks for information processing generated by an optical encoder based on the polarization of light. A machine learning technique is proposed to train the neural networks, such that the system can predict with remarkable accuracy the angular position in which the rotating element of the neuro-encoder is located, based on information provided by light’s phase shifting arrangements. The proposed neural designs show excellent performance in small angular intervals, and a methodology is proposed to avoid losing this remarkable characteristic in measurements from 0 to 180o or even to 360o. The neuro-encoder is implemented in simulation stage to obtain performance results. This study can be useful to improve capabilities of resolvers or other polyphasic sensors.

In the era of big data, feature engineering has proved its efficiency and importance in dimensionality reduction and useful information extraction from original features. Feature engineering can be expressed as dimensionality reduction and is divided into two types of methods such as feature selection and feature extraction. Each method has its pros and cons. There are a lot of studies to combine these methods. Sparse autoencoder (SAE) is a representative deep feature learning method that combines feature selection with feature extraction. However, existing SAEs do not consider the feature importance during training. It causes extracting irrelevant information. In this paper, we propose a parallel guiding sparse autoencoder (PGSAE) to guide the information by two parallel guiding layers and sparsity constraints. The parallel guiding layers keep the main distribution using Wasserstein distance which is a metric of distribution difference, and it suppresses the leverage of guiding features to prevent overfitting. We perform our experiments using four datasets that have different dimensionality and number of samples. The proposed PGSAE method produces a better classification performance compared to other dimensionality reduction methods.

This study introduces a novel method for early prediction of Kimchi cabbage (Brassica rapa subsp. pekinensis (Lour.) Hanelt) height, utilizing drone imagery and a long short-term memory (LSTM) model. The research was conducted on a testbed at the National Institute of Agricultural Sciences (NAS) in South Korea, encompassing two distinct soil types (loam and sandy loam) to investigate their impact on growth. High-resolution drone images were captured throughout the growing season to generate a canopy height model (CHM) for estimating plant height at various stages. Missing height data were interpolated using a logistic growth curve, and an LSTM model was trained on this data to predict the final height of Kimchi cabbage at harvest. Three LSTM models were developed using time-series data collected at 29, 36, and 44 days after planting (DAP). The model trained on data from DAP 44 demonstrated the highest accuracy with a coefficient of determination (R²) of 0.83, a mean absolute error (MAE) of 2.48 cm, and a root mean square error (RMSE) of 3.26 cm, outperforming models trained on earlier data. Color-coded maps were generated to visualize the spatial distribution of predicted Kimchi cabbage heights, revealing variations in growth patterns across the testbed and confirming the model's potential for site-specific management. Considering the trade-off between accuracy and prediction timing, the model trained on DAP 36 data (MAE = 2.77 cm) was deemed optimal for informing cultivation management decisions. This research demonstrates the feasibility and effectiveness of integrating drone imagery, logistic growth curves, and LSTM models for early and accurate prediction of Kimchi cabbage height. The proposed technology enables data-driven decision-making for farmers, facilitating timely interventions based on predicted growth patterns. This could lead to improved crop yields, resource optimization, and a more sustainable agricultural future. Future research will focus on refining the model's accuracy and exploring its applicability to other crops, further expanding the potential of precision agriculture technologies.

Semantic segmentation of remote sensing urban scene imagery is a pixel-wise prediction, which is applied to identify the land-cover or land-use category. However, semantic segmentation demands huge computation cost. In order to reduce the huge computation cost, a common method is to introduce transformer and CNN hybrid method to have a good trade-off between accuracy and computation cost. However, recent CNN-transformer hybrid methods often capture local-global context and cross-window interaction information simultaneously. Then they introduce the local-global context to fuse the local feature, which repeat fuse the local and global feature and result in the additional computation cost. And previous methods often ignore to filter the local and global feature. To fix the problem, we design a lightweight decoder-EDDformer, which is Efficient Global Value Transformer with Dynamic Gatefusion module. Efficient global value transformer only response for extracting global feature. Dynamic Gatefusion module filter the local and global semantic feature and fuse them to capture the local-global context in one time. Extensive experiments reveal that our method not only runs faster but also achieves higher accuracy compared with other state-of-the-art lightweight models.

In recent years, electroencephalogram (EEG) emotion recognition has emerged as a research hotspot in the field of artificial intelligence (AI). Despite advancements in EEG-based emotion recognition, room for enhancing performance persists due to EEG signal redundancy and limitations in feature extraction, leading to inefficient models and emotional information loss. To fully utilize EEG's emotional information and improve recognition accuracy while reducing computational costs, this paper proposes a Convolutional-Recurrent Hybrid Network with a dual-stream adaptive approach and an attention mechanism (CSA-SA-CRTNN). Firstly, the model utilizes a CSAM module to assign corresponding weights to EEG channels. Then, an adaptive dual-stream convolutional-recurrent network (SA-CRNN and MHSA-CRNN) is applied to extract local spatial-temporal features. After that, the extracted local features are concatenated and fed into a temporal convolutional network with a multi-head self-attention mechanism (MHSA-TCN) to capture global information. Finally, the extracted EEG information is used for emotion classification. We conducted binary and ternary classification experiments on the DEAP dataset, achieving 99.26% and 99.15% accuracy for arousal and valence in binary classification, and 97.69% and 98.05% in ternary classification, surpassing relevant algorithms. Additionally, the model's efficiency is significantly higher than other models, achieving better accuracy with lower resource consumption.

Accurate classification of objects in 3D point clouds is a significant problem in several applications, such as autonomous navigation and augmented/virtual reality scenarios, which has become a research hot spot. In this paper, we presented a deep learning strategy for 3D object classification in augmented reality. The proposed approach is a combination of the GRU and LSTM. LSTM networks learn longer dependencies well, but due to the number of gates, it takes longer to train; on the other hand, GRU networks have a weaker performance than LSTM, but their training speed is much higher than GRU, which is The speed is due to its fewer gates. The proposed approach used the combination of speed and accuracy of these two networks. The proposed approach achieved an accuracy of 0.99 in the 4,499,0641 points dataset, which includes eight classes (unlabeled, man-made terrain, natural terrain, high vegetation, low vegetation, buildings, hardscape, scanning artifacts, cars). Meanwhile, the traditional machine learning approaches could achieve a maximum accuracy of 0.9489 in the best case.

In today’s rapidly evolving digital landscape, the significance of password guessing techniques in both offensive and defensive strategies is paramount. Passwords serve as a crucial line of defense for both individuals and corporations, safeguarding their sensitive systems and data. Therefore, assessing the effectiveness of these access credentials is a critical task. However, existing research in this field often encounters limitations, such as a lack of sufficient training data and extended model training times. Current methods often struggle with limited training data and lengthy training times. This paper introduces BBPE-AE, a novel Auto Encoder Network (AE) designed for password guessing. BBPE-AE utilizes Byte-level Byte Pair Encoding (BBPE) to extract frequent tokens from password datasets without length restrictions, employing a dynamic window technique to capture complex patterns. Experimental results on Hotmail and Myspace datasets demonstrate exceptional performance, achieving high similarity rates (BLEU-Unigram: 0.90, BLEU-Bigram: 0.82 for Hotmail; BLEU-Unigram: 0.90, BLEU-Bigram: 0.81 for Myspace). BBPE-AE generates realistic passwords that meet HIBP (Have I Been Pwned) standards, minimizing duplication. These findings highlight the effectiveness of BBPE-AE in enhancing security by generating realistic passwords, ultimately safeguarding sensitive systems and data.

Bearing fault diagnosis is crucial for ensuring the stable operation of mechanical equipment. With the continuous development of deep learning technology, Convolutional Neural Networks (CNNs) have demonstrated significant advantages in the field of fault diagnosis. This paper proposes a new method that combines various CNN architectures to improve the accuracy of bearing fault diagnosis. We designed five different convolutional network structures, including SerConv, ResConv, One-Shot Aggregation Convolution (OSAConv), Cross-Stage Aggregation Convolution (CSAConv), and MD-DAConv. Experimental results on the Case Western Reserve University (CWRU) bearing dataset show that the proposed method exhibits high accuracy and robustness in fault diagnosis. The results indicate that strategies such as multi-directional, multi-scale, and residual connections play a crucial role in enhancing the depth and breadth of feature extraction, while simple and effective feature fusion and information transmission mechanisms are key to ensuring the robustness and generalization ability of the model.

Deep convolutional neural networks (CNNs) have revolutionised computer vision technology by automatically learning hierarchical representations of image data, leading to state-of-the-art performance in visual recognition tasks. This article presents a comprehensive review of deep CNNs, from their evolution and architectures to the current state-of-the-art research. The review also provides the core concepts and building blocks of CNNs and their concise mathematical representations. Furthermore, it explores applications of deep CNNs in three popular domains, including medical diagnosis, remote sensing, and facial recognition. This review will benefit researchers and practitioners in computer vision and artificial intelligence, as well as industry professionals seeking to leverage the latest advancements in deep learning technologies.

In a complex working environment, the fault signal of the gearbox is greatly affected by the outside world, and fault feature recognition is difficult, so the fault diagnosis accuracy is difficult to meet the expected requirements. To solve this problem, this paper proposes a gearbox fault diagnosis method based on an optimized stacked denoising auto encoder (SDAE) and kernel extreme learning machine (KELM). Firstly, the Particle Swarm Optimization algorithm in Adaptive Weight (SAPSO) was adopted to optimize the SDAE network structure, and the number of hidden layer nodes, learning rate, noise addition ratio and iteration times were adaptively obtained to make SDAE obtain the best network structure. Then, the best SDAE network structure was used to extract the deep feature information of weak faults in the original signal. Finally, the extracted fault features are fed into KELM for fault classification. Experimental results show that, compared with existing commonly used methods of fault diagnosis, the fault diagnosis model proposed in this paper can reduce the influence of noise in the original signal can better learn the deep-level features in the original signal and has higher diagnosis accuracy, faster diagnosis speed and good generalization in fault diagnosis.

This paper proposes two max-pooling engines, named the RTB-MAXP engine and the CMB-MAXP engine, with a scalable window size parameter for FPGA-based convolutional neural network (CNN) implementation. The max-pooling operation for the CNN can be decomposed into two stages, i.e., a horizontal axis max-pooling operation and a vertical axis max-pooling operation. These two one-dimensional max-pooling operations are performed by tracking the rank of the values within the window in the RTB-MAXP engine and cascading the maximum operations of the values in CMB-MAXP engine. Both the RBM-MAXP engine and the CMB-MAXP engine were implemented using VHSIC Hardware Description Language (VHDL) and verified by simulations. They have been employed for and tested in our CNN accelerator targeting at the CNN model YOLOv4-CSP-S-Leaky for object detection.

Cognitive neuroscience is the study of how the human brain functions on tasks like decision making, language, perception and reasoning. Deep learning is a class of machine learning algorithms that use neural networks. They are designed to model the responses of neurons in the human brain. Learning can be supervised or unsupervised. Ngram token models are used extensively in language prediction. Ngrams are probabilistic models that are used in predicting the next word or token. They are a statistical model of word sequences or tokens and are called Language Models or Lms. Ngrams are essential in creating language prediction models. We are exploring a broader sandbox ecosystems enabling for AI. Specifically, around Deep learning applications on unstructured content form on the web.

This study explores novel techniques to improve the detection accuracy of skarn iron deposits using advanced image processing methodologies. Leveraging the capabilities of Aster image, Band Ratio (BR) images, and Principal Component Analysis (PCA), alongside the power of 3D Convolutional Neural Networks (3D-CNN), the research aims to enhance the precision and efficiency of ore detection in complex geological environments. The proposed method employs a specific 3D-CNN architecture, accepting input as a 7×7×C image patch, where C represents the combined number of selected Aster image bands, Principal Component (PC) bands, and computed BR images. To evaluate the accuracy of the proposed method, five distinct image bands combinations, including the proposed bands combination, were tested and evaluated based on the overall accuracy (OA), average accuracy (AA), and kappa coefficient. The results demonstrated that while the incorporation of BR images alongside Aster bands initially seemed promising, it introduced significant confusion in certain classifications, leading to unexpected mis-classification rates. Surprisingly, utilizing solely Aster bands as input parameters yielded higher accuracy rates (OA= 93.13%, AA = 91.96%, Kappa = 90.91%) compared to scenarios involving the integration with band ratios (OA = 87.02%, AA = 79.15, and Kappa = 82.60%) or the integration of BR images to PC bands (OA= 87.78%, AA = 82.39%, Kappa = 83.81%). However, the amalgamation of Aster bands with selected PC bands showed slight improvements in accuracy (OA= 94.65%, AA = 92.93%, Kappa = 93.45%), although challenges in accurately classifying certain features persisted. Ultimately, the proposed combination of Aster bands, PC bands, and BR images (proposed bands combination) presented the most visually appealing and statistically accurate results (OA = 96.95%, AA= 94.87%, and Kappa = 95.93%), effectively addressing misclassifications observed in the other combinations. These findings underscore the synergistic contributions of each of the Aster bands, PC bands, and BR images, with Aster bands proving pivotal for optimal skarn classification, PC bands enhancing intrusions classification accuracy, and BR images strengthening wall rock classification accuracy. In conclusion, the proposed combination of input image bands emerges as a robust and comprehensive methodology, demonstrating unparalleled accuracy in the remote sensing detection of skarn iron minerals.

This article presents the implementation of a model to estimate the future price of commodities in the Brazilian market from time series of short-term technical evaluation. For this, data from two databases were used, one referring to the foreign market (opening values, maximum, minimum, closing, closing adjustment and volume) and the other, from the Brazilian market (the price of the day), considering commodities, sugar, cotton, corn, soybean and wheat. Subsequently, the technical indicators were calculated from the TA-Lib technical analysis library. Pearson’s correlation coefficient was applied, records with low correlation were removed, and then the database was consolidated. From the pre-processed data, Long Short-Term Memory (LSTM) recurrent neural networks were used to perform data prediction at the one and three day interval. These models were evaluated using the mean square error (MSE), obtaining results between 0.00010 and 0.00037 on test data one day ahead, and from 0.00017 to 0.00042 three days ahead. However, based on the results obtained, it was observed that the developed model obtained a promising forecasting performance for all the commodities evaluated. As a main contribution, there is the consolidation of databases that can be used in future scientific research. Furthermore, based on its interpretation, it can assist in decision making regarding the buying and selling of commodities to increase financial gains.

Sign language recognition technology can help people with hearing impairments to communicate with those who are hearing impaired. At present, with the rapid development of society, deep learning technology also provided certain technical support for sign language recognition work. In sign language recognition tasks, the use of traditional convolutional neural networks to extract spatio-temporal features from sign language videos suffers from insufficient feature extraction, resulting in low recognition rates. Nevertheless, video-based sign language datasets are very large and require a lot of computational resources for training and generalisation must be ensured, which poses a challenge for recognition. In this paper, we have presented a video-based sign language recognition method based on resnet and lstm. As the number of network layers increases, the ResNet network can effectively solve the granularity explosion problem and obtain better time series features. We use the Resnet convolutional network as the backbone model. At the initialisation stage, we obtain sign language features using ResNet; then, the learned feature space is used as the input of LSTM network to obtain long sequence features. The experimental results show that the accuracy of the above model is better than the mainstream model, and it can effectively extract the spatio-temporal features in sign language videos and improve the recognition rate of sign language actions.

We investigate the power of time series analysis based on a variety of information-theoretic approaches from statistics (AIC, BIC) and machine learning (Minimum Message Length) - and we then compare their efficacy with traditional time series model and with hybrids involving deep learning. More specifically, we develop AIC, BIC and Minimum Message Length (MML) ARMA (autoregressive moving average) time series models - with this Bayesian information-theoretic MML ARMA modelling already being new work. We then study deep learning based algorithms in time series forecasting, using Long Short Term Memory (LSTM), and we then combine this with the ARMA modelling to produce a hybrid ARMA-LSTM prediction. Part of the purpose of the use of LSTM is to seek capture any hidden information in the residuals left from the traditional ARMA model. We show that MML not only outperforms earlier statistical approaches to ARMA modelling, but we further show that the hybrid MML ARMA-LSTM models outperform both ARMA models and LSTM models.

Parkinson’s Disease (PD) is a neurodegenerative and progressive disease that impacts the nerve cells in the brain and varies from person to person. The exact cause of PD is still unknown, and the diagnosis of PD does not include a specific objective test with certainty. Although deep learning has made great progress in medical neuroimaging analysis, these methods are very susceptible to biases present in neuroimaging datasets. An innovative decorrelated deep learning technique is introduced to mitigate class bias and scanner bias while simultaneously focusing on finding distinguishing characteristics in resting-state functional MRI (rs-fMRI) data, which assist in recognizing the PD with good accuracy. The decorrelation function reduces the non-linear correlation between features and bias in order to learn bias-invariant features. The Parkinson’s Progression Markers Initiative (PPMI) dataset, referred to as a single scanner imbalanced dataset in this study used to validate our method. The imbalanced dataset problem affects the performance of the deep learning framework by overfitting to the majority class. To resolve this problem, we propose a new Decorrelated Convolutional Neural Networks (DcCNN) framework by applying decorrelation-based optimization to Convolutional Neural Networks(CNN). An analysis of evaluation metrics comparisons shows that integrating the decorrelation function boosts the performance of PD recognition by removing class bias. Specifically, our DcCNN model performs significantly better than existing traditional approaches to tackle the imbalance problem. Finally, the same framework can be extended to create scanner invariant features without significantly impacting the performance of a model. The obtained dataset is a multi-scanner dataset which leads to scanner bias due to the differences in acquisition protocols and scanners. The multi-scanner dataset is a combination of two datasets, namely PPMI and FTLDNI - frontotemporal lobar degeneration neuroimaging initiative (NIFD) dataset. The results of t-distributed stochastic neighbor embedding (t-SNE) and scanner classification accuracy of our proposed Feature Extraction-DcCNN (FE-DcCNN) model validated the effective removal of scanner bias. Our method achieves an average accuracy of 77.80% on a multi-scanner dataset for differentiating PD from healthy control, which is superior to the DcCNN model trained on a single scanner imbalanced dataset.

The detection and recognition of tables in image-based documents is a challenging and essential task for automated data extraction and anal- ysis. This paper introduces a methodology that utilizes Convolutional Neural Networks (CNNs) to address this problem. By harnessing the ro- bust visual recognition abilities of CNNs, the proposed method effectively detects and extracts tables from image-based documents. Additionally, an Encoder-Dual Decoder (EDD) architecture is employed to extract the textual content from table cells, enabling the transformation of visual data into structured, machine-readable formats. This structured data is then further processed using a Vector Space Model (VSM)-based language modeling technique for query-based table retrieval. The methodology’s accuracy is assessed by analyzing the spatial relationships between the extracted cells and comparing them against a pre-defined table structure. The results provide an evidence to the effectiveness of the proposed ap- proach, with an impressive 85% accuracy rate and a corresponding error rate of 15%. This accurate extraction and recognition of tables from image-based documents can significantly enhance applications in data analysis, information retrieval, and knowledge extraction.

The demand for precise indoor localization services is steadily increasing. Among various methods, fingerprint-based indoor localization has become a popular choice due to its exceptional accuracy, cost-effectiveness, and ease of implementation. However, its performance degrades significantly as a result of multipath signal attenuation and environmental changes. In this paper, we propose an indoor localization method based on fingerprints using self-attention and long short-term memory (LSTM). By integrating a self-attention mechanism and LSTM network, the proposed method exhibits outstanding positioning accuracy and robustness in diverse experimental environments. The performance of the proposed method is evaluated under two different experimental scenarios, which involve 2D and 3D moving trajectories, respectively. The experimental results demonstrate that our approach achieves an average localization error of 1.76 m and 2.83 m in the respective scenarios, outperforming the existing state-of-the-art methods by 42.67% and 31.64%.

Aiming at solving difficulties related to aero-engine classification and identification, the infrared spectra of aero-engine hot jets are measured by telemetry Fourier transform infrared spectrometer, and a six types aero-engine hot jet spectrum data set is generated. The spectrum data set were randomly sampled at a ratio of 8:1:1 to generate training set, validation set and prediction set. In this paper, a Peak Finding Siamese Convolutional Neural Network (PF-SCNN) is designed to match and classify spectrum data. In the first place, the matching pairs of training set and validation set are made independently with the labels of positive and negative samples creating. The spectrum feature extraction and distance similarity calculation are carried out by the Siamese convolutional neural network (SCNN). When entering the prediction phase, the prediction set and training set are integrated to predict in the trained model. The prediction set labels are ultimately determined by matching them with the training set labels having the highest similarities to ensure accurate predictions. To improve the operation efficiency of the SCNN, a peak finding method is introduced to extract the spectrum peaks. Peak positions of high frequency are counted, while the peak data are extracted. The peak data are used to take the place of the original data for training and prediction. The performance measures are defined as Accuracy, Precision, Recall, Confusion matrix and F1-score, and the prediction accuracy was up to 99%. The experimental results indicate that the proposed PF-SCNN is suitable for our spectral dataset and can complete the task of infrared spectrum classification of aero-engine hot jets.

Antimony (Sb) has gained significance as a critical raw material (CRM) within the European Union (EU) due to its strategic importance in various industrial sectors, particularly in the textile industry for flame retardants and as a component of Sb-based semiconductor materials. Moreover, Sb is emerging as a potential alternative for anodes used in lithium-ion batteries, a key element in the Energy transition. This study focused on exploring the feasibility of identifying and quantifying Sb mineralizations through the spectral signature of soils using reflectance spectroscopy, a non-invasive remote sensing technique, and by employing deep learning algorithms such as Convolutional Neural Networks (CNNs). Common signal preprocessing techniques were applied to the spectral data, and the soils were analyzed by Inductively Coupled Plasma Mass Spectrometry (ICP-MS). Despite achieving high R-squared values, the study faces a significant challenge of generalization of the model to new data. Despite the limitations, this study provides valuable insights into potential strategies for future research in this field.

Breast Cancer diagnosis is one of the most studied problems in the medical domain. In the medical domain, cancer diagnosis has been studied extensively which instantiates the need of early prediction of cancer disease. For obtaining advance prediction, health records are exploited and given as input to an automated system. This paper focuses on constructing an automated system by employing deep learning based recurrent neural network models. A stacked GRU-LSTM-BRNN is proposed in this paper that accepts health records of a patient for determining possibility of being affected by breast cancer. Proposed model is compared against other baseline classifiers such as stacked Simple-RNN model, stacked LSTM-RNN model, stacked GRU-RNN model. Comparative results obtained in this study indicate that stacked GRU-LSTM-BRNN yield better classification performance for predictions related to breast cancer disease.

With an explosion of data in the nearly every domain, from physical and environmental sciences to econometrics and even social science, we have entered a golden era of time-domain data science. In addition to ever increasing data volumes and varieties of time-series, there's a plethora of techniques available to analyses them. Unsurprisingly, machine learning approaches are very popular. In environmental sciences (ranging from climate studies to space weather/solar physics), machine learning (ML )and deep learning (DL) approaches complement the traditional data assimilation approaches. In solar and astrophysics, multiwavelength time-series or lightcurves along with observations of other messengers like gravitational waves provide a treasure trove of time-domain data for ML/DL approaches for modeling and forecasting. In this regard, deep learning algorithms such as recurrent neural networks (RNNs) should prove extremely powerful for forecasting as it has in several other domains. Such methods agnostic of the physics can only predict transient flares that are represented in the original distribution trained on (eg. coloured noise fluctuations) and not special, isolated episodes (eg. a one-off outburst like a major coronal mass ejection) potentially distinguishing them. Therefore it is vital to test robustness of these techniques ; here we attempt to make a small contribution to this by exploring sensitivity of RNNs with some rudimentary experiments. We test the performance of a very successful class of RNNs, the Long Short Term Memory (LSTM) algorithms with simulated lightcurves. As an example, we focus on univariate forecasting of time-series (or lightcurves) characteristic certain variable astrophysical objects (eg. active galactic nuclei or AGN). Specifically, we explore the sensitivity of training and test losses to key parameters of the LSTM network and data characteristics namely gaps and complexity measured in terms of number of Fourier components. We find that typically, the performances of LSTMs are better for pink or flicker noise type sources. The key parameters on which performance is dependent are batch size for LSTM and the gap percentage of the lightcurves. While a batch size of $10-30$ seems optimal, the most optimal test and train losses are under $10 \%$ of missing data for both periodic and random gaps in pink noise. The performance is far worse for red noise. This compromises detectability of transients. Thus, we show that generative models (time-series simulations) are excellent guides for use of RNN-LSTMs in building forecasting systems

Functional connectivity (FC) is a potential candidate that can increase the performance of brain-computer interfaces (BCIs) in the elderly because of its compensatory role in neural circuits. However, it is difficult to decode FC by current machine learning techniques because of a lack of its physiological understanding. To investigate the suitability of FC in BCI for the elderly, we propose the decoding of lower- and higher-order FCs using a convolutional neural network (CNN) in six cognitive-motor tasks. The layer-wise relevance propagation (LRP) method describes how age-related changes in FCs impact BCI applications for the elderly compared to younger adults. Seventeen younger (24.5±2.7 years) and twelve older (72.5±3.2 years) adults were recruited to perform tasks related to hand-force control with or without mental calculation. CNN yielded a six-class classification accuracy of 75.3% in the elderly, exceeding the 70.7% accuracy for the younger adults. In the elderly, the proposed method increases the classification accuracy by 88.3% compared to the filter-bank common spatial pattern (FBCSP). LRP results revealed that both lower- and higher-order FCs were dominantly overactivated in the prefrontal lobe depending on task type. These findings suggest a promising application of multi-order FC with deep learning on BCI systems for the elderly.

This study examines the feasibility of using event-related potentials (ERPs) obtained from electroencephalographic (EEG) recordings as biomarkers for long-term memory item classification. Previous studies have identified old/new effects in memory paradigms associated with explicit long-term memory and familiarity. Recent advancements in convolutional neural networks (CNNs) have enabled the classification of ERP trials under different conditions and identification of features related to neural processes at the single-trial level. We employed this approach to compare three CNN models with distinct architectures using experimental data. Participants (N = 25) performed an association memory task while recording ERPs that were used for training and validation of the CNN models. The EEGNET-based model achieved the most reliable performance in terms of precision, recall, and specificity compared with the shallow and deep convolutional approaches. The classification accuracy of this model reached 70% for known items and 62% for unknown items. Good overall accuracy requires a trade-off between recall and specificity and depends on the architecture of the model and the dataset size. These results suggest the possibility of integrating ERP and CNN into online learning tools and identifying the underlying processes related to long-term memorization.

In recent years, there has been a growing interest in remote sensing image-text cross-modal retrieval due to the rapid development of space information technology and the significant increase in remote sensing image data volume. One approach that has shown promising results in cross-modal retrieval of natural images is the multimodal fusion encoding method. However, remote sensing images have unique characteristics that make the retrieval task challenging. Firstly, the semantic features of remote sensing images are fine-grained, meaning they can be divided into multiple basic units of semantic expression. Additionally, these images exhibit variations in resolution, color, and perspective. Different combinations of basic units of semantic expression can generate diverse text descriptions. These characteristics pose considerable challenges for cross-modal retrieval. To address these challenges, this paper proposes a multi-task guided fusion encoder (MTGFE) based on the multimodal fusion encoding method. The model incorporates three tasks: image-text matching (ITM), masked language modeling (MLM), and the newly introduced multi-view joint representations contrast (MVJRC) task. By jointly training the model with these tasks, we aim to enhance its capability to capture fine-grained correlations between remote sensing images and texts. Specifically, the MVJRC task is designed to improve the model’s consistency in feature expression and fine-grained correlation, particularly for remote sensing images with significant differences in resolution, color, and angle. Furthermore, to address the computational complexity associated with large-scale fusion models and improve retrieval efficiency, this paper proposes a retrieval filtering method. This method achieves higher retrieval efficiency while minimizing accuracy loss. Extensive experiments were conducted on four public datasets to evaluate the proposed method, and the results validate its effectiveness. Overall, this study focuses on remote sensing image-text cross-modal retrieval and introduces the MTGFE model, which combines multimodal fusion encoding with multiple tasks to enhance the model’s ability to capture fine-grained correlations. Additionally, a retrieval filtering method is proposed to improve retrieval efficiency. Experimental results demonstrate the effectiveness of the proposed method.

Fine-grained image classification remains an ongoing challenge in the computer vision field, which is particularly intended to identify objects within sub-categories. It is a difficult task since there is a minimal and substantial intra-class variance. The current methods address the issue by first locating selective regions with Region Proposal Networks (RPN), object localization, or part localization, followed by implementing a CNN Network or SVM classifier to those selective regions. This approach, however, makes the process simple by implementing a single-stage end-to-end feature encoding with a localization method, which leads to improved feature representations of individual tokens/regions by integrating the transformer encoder blocks into the Yolov5 backbone structure. These Transformer Encoder Blocks, with their self-attention mechanism, effectively captured the global dependencies and enabled the model to learn relationships between distant regions. This improved the model ability to understand context and captured long-range spatial relationships in the image. We also replaced the Yolov5 detection heads with three transformer heads at the output for object recognition using the discriminative and informative features maps from transformer encoder blocks. We established the potential of the single stage detector for the fine-grained image recognition task, by achieving state of the art 93.4% accuracy, as well as outperforming the existing Yolov5 model. The effectiveness of our approach is assessed using the Stanford car dataset, which includes 16,185 images of 196 different classes of vehicles with significantly identical visual appearances.

Measuring pulmonary nodules accurately can help with early diagnosis of lung cancer, which can improve a patient’s chances of survival. Many methods for segmenting nodules have been developed, but they all rely on input from radiologists in the form of a 3D volume of interest (VOI) or use a constant region of interest (ROI) and only consider the presence of nodule within the given VOI. These approaches limit the networks’ ability to detect nodules outside the given VOI and can also include unnecessary structures in the VOI, leading to potentially inaccurate segmentation. In this work, we propose a novel approach for 3D lung nodule segmentation by using 2D region of interest (ROI) inputted from radiologist or computer-aided detection (CADe) system. Particularly, we design a dual-encoder-based hard attention network (DEHA-Net) which incorporates the full slice of thoracic computed tomography scan along with the ROI mask to produce an accurate segmentation mask of lung nodule in the given slice. The proposed architecture exploits the adaptive region of interest (A-ROI) algorithm to automatically investigates the penetration of lung nodule into surrounding slices while eliminating the need to drawing separate ROIs in each slice. Further, the framework performs the multi-view analysis, i.e., in sagittal and coronal views, to improve the segmentation performance. The proposed scheme has been rigorously evaluated on the lung image database consortium and image database resource initiative (LIDC/IDRI) dataset and an extensive analysis of results have been performed. The quantitative analysis shows that the proposed method not only improves the existing state-of-the-art in term of dice score but also, significantly robust against the different types, shape and dimensions of lung nodules.

Caring for dolphins is a delicate process that requires experienced caretakers to pay close attention to their behavioral characteristics. However, caretakers may sometimes lack experience or not be able to give their full attention, which can lead to misjudgment or oversight. To address these issues, a dolphin behavior analysis system has been designed to assist caretakers in making accurate assessments. The study utilized image preprocessing techniques to reduce sunlight reflection in the pool and enhance the outline of dolphins, making it easier to analyze their movements. The dolphins were divided into 11 keypoints using an open-source tool called DeepLabCut, which helped accurately mark various body parts for skeletal detection. The AquaAI Dolphin Decoder (ADD) was then used to analyze six dolphin behaviors. To improve behavior recognition accuracy, the Long Short-Term Memory (LSTM) neural network was introduced. The ADD and LSTM models were integrated to form the ADD-LSTM system. Several classification models, including unidirectional and bidirectional LSTM, GRU, and SVM, were compared. The results showed that the ADD module combined with a double-layer bidirectional LSTM method achieved high accuracy in dolphin behavior analysis. The accuracy rates for each behavior were over 90%.

Predicting the prices of cryptocurrency owing to its volatility, instability, and other factors has been challenging; investors and traders especially in Nigeria have been on a constant look for a more reliable way of knowing market trends and prices and while there has been so many research conducted using deep learning, the results for which has fall short of what investors could called a strong predictor. This research reviewed the results of many works that had been done and proposed two types of recurrent neural network (RNNs) namely Long Short-Term Memory and Gated Recurrent Unit (GRU) for predicting the future prices of two of the most common crypto assets namely Bitcoin (BTC) and Ethereum (ETH), these two were selected based on their popularity, the volume traded and their market capitalization. The experiment was conducted in a GPU Jupyter Notebook environment and the performance of our experiment was evaluated on a test set using root mean square error (RMSE) and based on its values, the LSTM presented a better performance with rsme scores of 654.66, and 80.30 respectively for BTC and ETH as compared to GRU. The paper proceeds further to compare the results of this experiment with other related works and we discovered that its performance is a great improvement.

Coastal wave modeling and forecasting are essential in oceanography, sustainable marine energy, and ocean engineering. Precise forecasting of wave speed and direction are crucial for offshore operations, marine energy, risk management, environmental management, coastal and sustainable maritime management. The coastline of Brunei Darussalam can generate between 15 and 126 Giga Watt of wave energy. In this experimental research, we used two numerical approaches, the finite difference, and spectral element methods, to model and simulate wave speed and direction. We calculated the mean error between numerical and analytical solutions. We proposed a novel, promising, univariate time series forecasting model by combining the Long Short-Term (LSTM) with KerasTuner hyperparameters tuning and optimization techniques. This method helps us to improve the accuracy and efficiency of time-series forecasting.. The experimental data was computed from high-precision Acoustic Doppler Current Profiler (ADCP) sensor data. This research is part of the preliminary feasibility analysis of wave energy production in Brunei Darussalam and net zero commitment for a sustainable environment. Seven independent forecast experiments were performed for wave speed and direction in degree and radian units for 1, 3, 6, 8, 10, 12, and 24 hours. Mean squared error (MSE) was adopted as a metric for both training and testing. The experimental results reveal that the wave speed forecast has the lowest MSE compared to direction, regardless of the unit of measure, but has a longer duration. In addition, the direction forecast in the degree unit has the lowest errors compared to the unit of radians; the latter has a longer running time than the former. The model has delivered optimal results throughout the experiments with minor training and test errors. We conducted a thorough evaluations on two benchmark time series datasets, which include the study dataset and air quality index dataset, to validate the performance of the proposed model with other models. The proposed model outperforms cutting-edge forecasting models, such as the conventional LSTM, ARIMA, and Prophet. The model has the slightest forecast error compared to the existing literature’s result.

The choice of holiday destinations is highly depended on climate considerations. Nowadays, since the effects of climate crisis are being increasingly felt, the need of accurate weather and climate services for hotels is crucial. Such a service could be beneficial for both the future planning of tourists’ activities and destinations and for hotel managers as it could help in decision making about the planning and expansion of the touristic season, due to a prediction of higher temperatures for a longer time span, thus causing increased revenue for companies in the local touristic sector. The aim of this work is to calculate predictions on climatic variables using statistical techniques as well as Artificial Intelligence (AI) for a specific area of interest utilising data from in situ meteorological station, and produce valuable and reliable localised predictions with the most cost-effective method possible. This investigation will answer the question of the most suitable prediction method for time series data from a single meteorological station that is deployed in a specific location. As a result, an accurate representation of the microclimate in a specific are is achieved. To achieve this high accuracy in situ measurements and prediction techniques are used. As prediction techniques, Seasonal Auto Regressive Integrated Moving Average (SARIMA), AI techniques like the Long-Short-Term-Memory (LSTM) Neural Network and hybrid combinations of the two are used. Variables of interest are divided in the easier to predict temperature and humidity that are more periodic and less chaotic, and the wind speed as an example of a more stochastic variable with no known seasonality and patterns. Our results show that the examined Hybrid methodology performs the best at temperature and wind speed forecasts, closely followed by the SARIMA whereas LSTM perform better overall at the humidity forecast, even after the correction of the Hybrid to the SARIMA model.

Stock price prediction is of great significance but faces numerous challenges. In this study, a hybrid CNN-LSTM model was utilized. The stock data were processed and feature engineering was conducted. The model architecture and training strategies were expounded. Experimental results demonstrated that this model outperformed traditional methods and benchmark models on the test set, featuring strong capabilities in feature extraction and handling long-term dependencies. Key roles of data preprocessing, hyperparameter adjustment, and model fusion were also summarized, providing valuable references for stock price prediction.

In this paper, a special recurrent neural network (RNN) called \emph{Long Short-Term Memory (LSTM)} is used to design a virtual load sensor that estimates the mass of heavy vehicles. The estimation algorithm consists of a two-layer LSTM network, with the two layers based on sequence-to-sequence and sequence-to-one logic, respectively. The network estimates the vehicle mass based on the vehicle speed, longitudinal acceleration, engine speed, engine torque, and the accelerator pedal position. The network is trained and tested with a data set collected in a high-fidelity simulation environment called Truckmaker. Training data is generated in acceleration maneuvers in a range of speeds whereas test data is obtained by simulating the vehicle in a Worldwide harmonized light vehicles test cycle (WLTC). Preliminary results show that, with the proposed approach, the heavy-vehicle mass can be estimated as accurately as commercial load sensors in a range of load mass as wide as four tons.

Estimating the Remaining Useful Life (RUL) of aircraft engines holds a pivotal role in enhancing safety, optimizing operations, and promoting sustainability, thus being a crucial component of modern aviation management. Precise RUL predictions offer valuable insights into an engine’s condition, enabling informed decisions regarding maintenance and crew scheduling. In this context, we propose a novel RUL prediction approach in this paper, harnessing the power of Bi-directional LSTM and Transformer architectures, known for their success in sequence modeling, such as natural languages. We adopt the encoder part of the full Transformer as the backbone of our framework, integrating it with a self-supervised denoising autoencoder that utilizes Bidirectional LSTM for improved feature extraction. Within our framework, a sequence of multivariate time series sensor measurements serves as the input, initially processed by the Bidirectional LSTM autoencoder to extract essential features. Subsequently, these feature values are fed into our Transformer encoder backbone for RUL prediction. Notably, our approach simultaneously trains the autoencoder and Transformer encoder, different from the naive sequential training method. Through a series of numerical experiments carried out on the C-MAPSS datasets, we demonstrate that the efficacy of our proposed models either surpasses or stands on par with that of other existing methods.

Fluctuating stock prices make it difficult for investors to see investment opportunities. One tool that can help investors overcome this is forecasting techniques. Long Short-Term Memory (LSTM) is one of deep learning methods used in forecasting time series. The training and success of deep learning is strongly influenced by the selection of hyperparameters. This research uses a hybrid method between the Genetic Algorithm (GA) and LSTM to find a suitable model for predicting stock prices. GA is used in optimizing the architecture such as the number of epochs, window size, and the number of LSTM units in the hidden layer. Tuning optimizer is also carried out using several optimizers to achieve the best value. From method that has been applied, it shows that the method has a good level of accuracy with MAPE values below 10% in every optimizer used. The error rate generated is quite low, in case-1 with a minimum RMSE value of 93.03 and 94.40, & in case-2 with an RMSE value of 104.99 and 150.06 during training and testing. A fairly stable and small value is generated by setting it using the Adam Optimizer.

Early diagnosis of paroxysmal atrial fibrillation (PAF) could suggest patients to receive timely interventions in clinical practice. Various PAF onset prediction algorithms might benefit from accurate heart rate variability (HRV) analysis and machine learning classification but to be challenged in real-time monitoring scenarios. The aim of this study is to present an end-to-end deep learning based PAFNet model that integrates a sliding window technique on raw R-R interval of electrocardiogram (ECG) segments to achieve a real-time prediction of PAF onsets. This integration enable the deep convolutional neural network (CNN) to be customized as a light-weight architecture that accommodates the size of sliding windows simply by altering the input layer, specifically its effectiveness in making a new prediction with each new heartbeat. Catering to potential influence of input sizes, three CNN models were trained using 50, 100, and 200 R-R intervals, respectively. For each model, the performance of automated algorithms was evaluated for PAF prediction using a ten-fold cross-validation. As a results, a total of 56,381 PAFN-type and 56,900 N-type R-R interval segments was collected from publicly accessible ECG databases, and a promising prediction performance of the automated algorithm with a 100 R-R intervals was achieved in terms of the sensitivity of 97.12%, the specificity of 97.77%, and the accuracy of 97.45%, respectively. Importantly, the automated algorithm with a sliding windows step of 1 could process one sample in only 23.1 milliseconds and identify the onset of PAF at least 45 minutes in advance. The present results suggest that the sliding window technique on raw R-R interval sequences along by deep learning based algorithms may offers the possibility of providing an accurate, real-time, and end-to-end clinical tool for mass monitoring of PAF.

Public transportation scheduling aims to optimize the allocation of transportation resources, enhance transportation efficiency, and increase passenger satisfaction, which is crucial for building a sustainable urban transportation system. As a complement to public transportation, shared bikes provide users with a solution for the last mile of travel, compensating for the lack of flexibility in public transportation and helping to improve its utilization rate. Due to the characteristics of shared bikes, such as peak usage periods in the morning and evening and significant fluctuations in demand across different areas, the optimization of shared bike dispatch can better meet user needs, thereby reducing vehicle vacancy rates and increasing operating revenue. Addressing this issue, this paper designs a comprehensive decision-making approach for spatio-temporal demand prediction and bike dispatch optimization. For demand prediction, we designed a T-GCN based bike demand prediction model. In terms of dispatch optimization, we considered factors such as dispatch capacity, distance restrictions, and dispatch costs, and designed an optimization solution based on genetic algorithms. Finally, this paper validates the approach using shared bike operating data, shows that T-GCN can effectively predict short-term demand for shared bikes. Meanwhile, the optimization model based on genetic algorithms provides a complete dispatch solution verifying the model’s effectiveness. The shared bike dispatch approach proposed in this paper, which combines demand prediction with resource scheduling. This scheme can also be extended to other transportation scheduling problems with uncertain demand, such as store replenishment delivery and intercity inventory dispatch.

Background: Accurate sales forecasting in the mobile phone industry is critical to informed decision making by both the government and corporations. This study seeks to enhance mobile phone sales predictions by incorporating brand exposure as a predictive factor, hence addressing the gap left by the existing forecasting models that do not consider the impact of brand awareness. Method: This study thereby indicates a quantification method for brand exposure that merges word-of-mouth reviews with online search data. Word2Vec is used to extract initial keywords from a word-of-mouth corpus, whereas time-lag correlation analysis refines the core keywords. Later, PCA synthesizes brand exposure. The attention-LSTM framework hereby develops a forecasting model of mobile phone sales, effectively combining long short-term memory networks with the ability to capture time series dependencies and an attention mechanism to reduce data redundancy. Results: By considering brand exposure, the experiments demonstrate that the attention-LSTM model decreases the RMSE and MAPE indicators by 2.02% and 0.96%, respectively. In addition, compared with the ARIMA, SVR, BP neural network, and LSTM models, the attention-LSTM model decreases the average percentage errors by 6.52%, 3.42%, 2.56%, and 0.81%, respectively. Conclusion: Attention-LSTM captures dynamic sales trends very well and holds great significance over traditional models. Overall, including brand exposure, this model is very strong for the interpretation of mobile phone sales predictions, which has important implications in terms of policy as well as corporate strategies.

This article comprehensively explores multiple aspects of cryptocurrencies and their price forecasting. Firstly, the article introduces the definition of cryptocurrency and its development process on a global scale, especially focusing on the launch of Facebook Libra and China's central bank digital currency, highlighting the importance and influence of digital currency in the global financial market. Subsequently, the article analyzes the advantages of digital currencies over traditional currencies, including improving economic transaction efficiency, reducing transaction costs and enhancing transaction transparency. Meanwhile, the article also explores the challenges and risks potentially brought by the development of digital currencies, such as regulatory uncertainty and market volatility. In this context, the article raises the importance of cryptocurrency price forecasting and introduces the forecasting models and techniques commonly used today. Finally, through specific experimental analysis, the effectiveness of using the deep learning model CNN-LSTM to predict the price of Bitcoin is demonstrated, and the directions of future research and optimization strategy are proposed. In summary, this paper comprehensively presents the research status and prospects of cryptocurrency and its price prediction field through systematic introduction and analysis.

Fire outbreaks continue to cause damage despite the improvements in fire-detection tools and algorithms. It is still challenging to implement a well performing and optimized approach, which is sufficiently accurate, and has tractable complexity and low false alarming rate. Small amount of fire and identification of fire from a long distance is also a challenge in previously proposed techniques. In this study, we propose a novel hybrid model based on Convolutional Neural Networks (CNN) to detect and analyze fire intensity. 21 convolutional layers, 24 Rectified Linear Unit (ReLU) layers, 6 pooling layers, 3 fully connected layers, 2 dropout layers, and a softmax layer are included in the proposed 57-layer CNN model. Our proposed model performs instance segmentation in order to distinguish between fire and non-fire events. To reduce the intricacy of the proposed model, we also propose a key-frame extraction algorithm. The proposed model uses Internet of Things (IoT) devices to alert the relevant person by calculating the severity of fire. Our proposed model is tested on a publicly available dataset having fire and normal videos. The achievement of 95.25 % classification accuracy, 0.09% False Positive Rate (FPR), 0.65 percent False Negative Rate (FNR), and a prediction time of 0.08 seconds validates the proposed system.

Coconut is an important multipurpose crop wherein different stages of its maturity are used for different products. Traditional approach of its classification use sound formed by tapping the fruit, and the tapper will base its judgment through the sound he heard. The procedure was highly subjective and attempts had been made by other researchers to automate and classify the adulthood of the coconut objectively. Nowadays, deep learning techniques are being utilized to solve classification problems. One such method is Convolutional Neural Network (CNN) that takes raw pixel data, learns to extract features, and ultimately classifies the input. In this study, a portable device was developed that takes audio signals generated from tapping the fruit mechanically. This audio clip was converted to spectrogram and was used as input to a LeNet CNN architecture. We argue that CNN classifier has improved accuracy compared to the non-deep learning system. Our evaluation confirmed that the use of CNN had improved the accuracy of classifying coconut by about 15%.

A convolutional neural network (CNN) is sometimes understood as a black box in the sense that while it can approximate any function, studying its structure will not give us any insights into the nature of the function being approximated. In other terms, the discriminative ability does not reveal much about the latent representation of a network. This research aims to establish a framework for interpreting the CNNs by profiling them in terms of interpretable visual concepts and verifying them by means of Integrated Gradient. We also ask the question, "Do different input classes have a relationship or are they unrelated?" For instance, could there be an overlapping set of highly active neurons to identify different classes? Could there be a set of neurons that are useful for one input class whereas misleading for a different one? Intuition answers these questions positively, implying the existence of a structured set of neurons inclined to a particular class. Knowing this structure has significant values; it provides a principled way for identifying redundancies across the classes. Here the interpretability profiling has been done by evaluating the correspondence between individual hidden neurons and a set of human-understandable visual semantic concepts. We also propose an integrated gradient-based class-specific relevance mapping approach that takes the spatial position of the region of interest in the input image. Our relevance score verifies the interpretability scores in terms of neurons tuned to a particular concept/class. Further, we perform network ablation and measure the performance of the network based on our approach.

This research investigates the application of deep learning in sentiment analysis of Canadian maritime case law. It offers a framework for improving maritime law and legal analytics policy-making procedures. The automation of legal document extraction takes center stage, underscoring the vital role sentiment analysis plays at the document level. Therefore, this study introduces a novel strategy for sentiment analysis in Canadian maritime case law, combining sentiment case law approaches with state-of-the-art deep learning techniques. The overarching goal is to systematically unearth hidden biases within case law and investigate their impact on legal outcomes. Employing Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) based models, this research achieves a remarkable accuracy of 98.05% for categorizing instances. In contrast, Conventional Machine Learning techniques such as Support Vector Machines (SVM) yield an accuracy rate of 52.57%, Naive Bayes at 57.44 %, and Logistic Regression at 61.86%. The superior accuracy of the CNN and STM model combination underscores its usefulness in legal sentiment analysis, offering promising future applications in diverse fields like legal analytics and policy design. These findings mark a significant choice for AI-powered legal tools, presenting more sophisticated and sentiment-aware options for the legal profession.

: Innovative solutions are now being researched to manage the ever-increasing amount of data required to optimize the performance of internal combustion engines. Machine Learning ap-proaches have shown to be a valuable tool for signal prediction due to their real-time and cost-effective deployment. Among them, the architecture consisting of Long Short-Term Memory (LSTM) and one-dimensional Convolutional Neural Networks (1DCNN) has emerged as a highly promising and effective option for replacing the role of physical sensors. The architecture com-bines the capacity of LSTM to detect patterns and relationships in smaller segments of the signal with the ability of 1DCNN to detect patterns and relationships in larger segments of the signal. The purpose of this work is to assess the feasibility of substituting a physical device dedicated to calculating the torque supplied by a spark-ignition engine. The suggested architecture was trained and tested using signals from the field during a test campaign conducted under transient operating conditions. The results reveal that LSTM+1DCNN is particularly well suited for signal prediction with considerable variability. It constantly outperforms other architectures used for comparison, with average error percentages of less than 2%, proving the architecture's ability to replace physical sensors.

Alzheimer's disease (AD) is a progressive and evolving neurodegenerative disease with an insidious onset that can lead to memory loss and cognitive impairment. There is no effective treatment for this disease. However, early diagnosis plays an important role in treatment planning to slow down its progression, as treatment has the greatest impact in the early stages of the disease. Neurological images obtained through different imaging techniques provide powerful information and help diagnose the disease. With the wide application of deep learning techniques in disease diagnosis, especially the prominence of Convolutional Neural Networks (CNNs) in computer vision and image processing, more and more studies are proposing the use of this algorithm for the diagnosis of AD. In this paper, we first systematically introduce the impact of AD on people, detailing the biomarkers, early clinical symptoms, and risk factors of this disease. Secondly, it goes on to detail the development of CNNs, their form, and methods to help diagnose AD. It is proposed that CNNs can help diagnose AD by analyzing medical imaging data, particularly structural brain scans such as magnetic resonance imaging (MRI) and functional scans such as positron emission tomography (PET). Finally, it is concluded that CNNs are of great importance for the diagnosis of AD and that they are likely to play an increasingly important role in the early detection of the disease, the understanding of disease mechanisms, and ultimately, in the development of effective AD therapies and interventions. CNNs are playing an increasingly important role in the Their potential impact on healthcare emphasizes the importance of continued research and innovation in neural networks and medical imaging.

Accurate and robust deep learning models for facial expression recognition are challenging to achieve, given the diversity of human faces and variations in images, including different facial poses and lighting conditions. In this work, we proposed a clock-driven convolutional Spiking Neural Network (SNN) and STDP fine-tune architecture, meticulously calibrated its hyperparameters, and experimented with various optimization methods. The best model resulted was trained and evaluated on the Fer2013 and FER+ database, obtaining an accuracy of 61.87% and 79.97% without requiring auxiliary training data or face registration. To our best knowledge, the proposed SNN achieved comparable accuracy to CNNs of similar depth and possessed the advantages of low energy consumption and high computational efficiency. The computational efficiency of the proposed SNN is approximately three times that of CNNs. Along with this, we introduced the very recent cumulative spike guided encoder visualization technique and revealed the strong encoding capability of the proposed SNN.

In the realm of computer vision, semantic segmentation is the task of recognizing objects in images at the pixel level. This is done by performing a classification of each pixel. The task is complex and requires sophisticated skills and knowledge about the context to identify objects’ boundaries. The importance of semantic segmentation in many domains is undisputed. In medical diagnostics, it simplifies the early detection of pathologies, thus mitigating the possible consequences. In this work, we provide a review of the literature on deep ensemble learning models for polyp segmentation and we develop new ensembles based on convolutional neural networks and transformers. The development of an effective ensemble entails ensuring diversity between its components. To this end, we combine different models (HarDNet-MSEG, Polyp-PVT, and HSNet) trained with different data augmentation techniques, optimization methods, and learning rates, which we experimentally demonstrate to be useful to form a better ensemble. Most importantly, we introduce a new method to obtain the segmentation mask which is more suitable for combining transformers in an ensemble. In our extensive experimental evaluation, the proposed ensembles exhibit state-of-the-art performance.

An application of deep convolutional neural network and recurrence plot for financial market movement prediction is presented. Though it is challenging and subjective to interpret its information, the pattern formed by a recurrence plot provide a useful insight into the dy- namical system. We used a recurrence plot of seven financial time series to train a deep neural network for financial market movement predic- tion. Our approach is tested on our dataset and achieved an average of 53.25% classification accuracy. The result suggests that a well trained deep convolutional neural network can learn a recurrence plot and pre- dict a financial market direction.

With the increase of the Artificial Neural Network (ANN), machine learning has taken a forceful twist in recent times. One of the most spectacular kinds of ANN design is the Convolutional Neural Network (CNN). The Convolutional Neural Network (CNN) is a technology that mixes artificial neural networks and up to date deep learning strategies. In deep learning, Convolutional Neural Network is at the center of spectacular advances. This artificial neural network has been applied to several image recognition tasks for decades and attracted the eye of the researchers of the many countries in recent years as the CNN has shown promising performances in several computer vision and machine learning tasks. This paper describes the underlying architecture and various applications of Convolutional Neural Network.

The escalating reliance on revolutionary online web services has introduced heightened security risks, with persistent challenges posed by phishing despite extensive security measures. Traditional phishing systems, reliant on machine learning and manual features, struggle with evolving tactics. Recent advances in deep learning offer promising avenues for tackling novel phishing challenges and malicious URLs. This paper introduces a two-phase stack generalized model named AntiPhishStack, designed to detect phishing sites. The model leverages the learning of URLs and character-level TF-IDF features symmetrically, enhancing its ability to combat emerging phishing threats. In Phase I, features are trained on a base machine learning classifier, employing K-fold cross-validation for robust mean prediction. Phase II employs a two-layered stacked-based LSTM network with five adaptive optimizers for dynamic compilation, ensuring premier prediction on these features. Additionally, the symmetrical predictions from both phases are optimized and integrated to train a meta-XGBoost classifier, contributing to a final robust prediction. This work's significance lies in advancing phishing detection with AntiPhishStack, operating without prior phishing-specific feature knowledge. Experimental validation on two benchmark datasets, comprising benign and phishing or malicious URLs, demonstrates the model's exceptional performance, achieving a notable 96.04% accuracy. This research adds value to the ongoing discourse on symmetry and asymmetry in information security and provides a forward-thinking solution for enhancing network security in the face of evolving cyber threats.

Cultural Tourism (CT), which enhances the economic development of a region, aids a country in reinforcing its identities, enhancing cross-cultural understanding, and preserving the heritage culture of an area. Designing a proper tourism model assists the tourists in understanding the point of interest without the help of a local guide. However, owing to the need for the analysis of different factors, designing such a model is a complex process. Therefore, this article proposes a CT model for Riyadh for peak visitor time. Primarily, the map data and cultural event dataset are processed for location, such as grouping with Kriging Interpolation-based Chameleon (KIC), tree forming, and feature extraction. After that, the event dataset’s attributes are processed with word embedding. Meanwhile, the Social Network Sites (SNS) data like reviews and news are extracted with an external Application Programming Interface (API). The review data is processed with keyword extraction and word embedding, whereas the news data is processed with score value estimation. Lastly, the data are fused corresponding to a historical site and given to the Multi-Quadratic-Long Short Term Memory (MQ-LSTM) Recommendation System (RS); also, the recommended result with the map is stored in a database. Lastly, the database security is maintained with Locality Sensitive Hashing (LSH); moreover, by attaining higher performance values, the proposed model is experimentally verified.

This paper introduces a multi-class hand gesture recognition model developed to identify a set of defined hand gesture sequences in two-dimensional RGB video recordings. The work presents an action detection classifier that looks at both appearance and spatiotemporal parameters of consecutive frames. The classifier utilizes a convolutional-based network combined with a long-short-term memory unit. To leverage the need for a large-scale dataset, the model uses an available dataset to then adopt a technique known as transfer learning to fine-tune the model on the hand gestures of relevance. Validation curves performed over a batch size of 64 indicate an accuracy of 93.95% (± 0.37) with a mean Jaccard index of 0.812 (± 0.105) for 22 participants. The presented model illustrates the possibility of training a model with a small set of data (113,410 fully labelled frames). The proposed pipeline embraces a small-sized architecture that could facilitate its adoption.

Speech emotion recognition remains a heavy lifting in natural language processing. It has strict requirements to the effectiveness of feature extraction and that of acoustic model. With that in mind, a Heterogeneous Parallel Convolution Bi-LSTM model is proposed to address these challenges. It consists of two heterogeneous branches: the left one contains two dense layers and a Bi-LSTM layer, while the right one contains a dense layer, a convolution layer, and a Bi-LSTM layer. It can exploit the spatiotemporal information more effectively, and achieves 84.65%, 79.67%, and 56.50% unweighted average recall on the benchmark databases EMODB, CASIA, and SAVEE, respectively. Compared with the previous research results, the proposed model achieves better performance stably.

Speech/music classification that facilitates optimized signal processing from classification results has been extensively adapted as an essential part of various electronics applications, such as multi-rate audio codecs, automatic speech recognition, and multimedia document indexing. In this paper, a new technique to improve the robustness of speech/music classifier for 3GPP enhanced voice service (EVS) using long short-term memory (LSTM) is proposed. For effective speech/music classification, feature vectors implemented with the LSTM are chosen from the features of the EVS. Experiments show that LSTM-based speech/music classification produces better results than conventional EVS under a variety of conditions and types of speech/music data.

This article focuses on generating an alternative in order to identify traffic officers during driving. This research employed the latest You Only Look Once (YOLO) model, using a six-phase methodology: data collection, data preparation involving resizing and labeling, implementation of various filters to avoid overfitting, model training, prediction evaluation, and result interpretation. The YOLO model was applied across three iterations using a dataset of 1862 images. To enhance training efficiency and detection speed, the graphics processing unit (GPU) acceleration was utilized, further enhancing the experimental process. The results of this study revealed that the YOLOv8x variant produced the most promising results. This proposed model attained a remarkable F1 score of 0.95, bolstered by a confidence score of 0.631, with the potential for an increase to 0.80 in confidence without significantly compromising the F1-score. These findings are poised to make a substantial contribution to the broader research landscape, particularly in advancing the effectiveness of detection models for traffic officers.

Convolutional neural network (CNN), a class of artificial neural network (ANN) is attracting interests of researchers in all research domain. CNN was invented for computer vision. They have also shown to be useful for semantic parsing, sentence modeling and other natural language processing related tasks. Here in this paper we discuss the basics of CNN models and their scope to provide a reference/baseline to the researchers interested in using CNN models in their research.

Many works have recently identified the need to combine deep learning with extreme learning to strike a performance balance with accuracy especially in the domain of multimedia applications. Considering this new paradigm, namely convolutional extreme learning machine (CELM), we present a systematic review that investigates alternative deep learning architectures that use extreme learning machine (ELM) for a faster training to solve problems based on image analysis. We detail each of the architectures found in the literature, application scenarios, benchmark datasets, main results, advantages, and present the open challenges for CELM. We follow a well structured methodology and establish relevant research questions that guide our findings. We hope that the observation and classification of such works can leverage the CELM research area providing a good starting point to cope with some of the current problems in the image-based computer vision analysis.

Anomaly occurrences in hydraulic machinery may lead to massive systems shut down, jeopardizing the safety of the machinery and its surrounding human operator(s) and environment, and the severe economic implications succeeding the faults and their associated damage. Hydraulics are mostly placed in ruthless environments, where they are consistently vulnerable to many faults. Hence, not only the machines and their components are prone to anomalies, but also the sensors attached to them, which monitor and report their health and behavioral changes. In this work, a comprehensive applicational analysis of anomalies in hydraulic systems extracted from a hydraulic test rig is thoroughly achieved. Firstly, we provided a combination of a new architecture of LSTM autoencoders and supervised machine and deep learning methodologies to perform two separate stages of fault detection and diagnosis. The two phases are condensed by: the detection phase using the LSTM autoencoder. Followed by the fault diagnosis phase represented by the classification schema. The previously mentioned framework is applied to both component and sensor faults in hydraulic systems, deployed in the form of two in-depth applicational experiments. Moreover, a thorough literature review of the past decade related work for the two stages separately is successfully conducted in this paper.

In this article, we introduce a new approach to human movement by defining the movement as a static object or a super object in one two-dimensional image. This method can allow researchers to label and describe the total movement as an object isolated from a reference video. This ap-proach allows us to perform various tasks, including finding similar movements in a video, measuring, and comparing movements, generating new similar movements, and defining chore-ography by controlling specific parameters in the human body skeleton. As a result of the pre-sented approach, we can eliminate the need to label images manually, disregard the problem of finding the beginning and the end of a movement, overcome synchronization issues between movements, and perform any deep learning network-based operation that processes super objects in images in general. As part of this article, we will demonstrate two application use cases: one il-lustrates how to verify and score a requested movement. In contrast, the other illustrates how to generate similar movements in the human skeleton space by addressing the challenge of supply-ing sufficient training data for deep learning applications (DL). A Variational Auto Encoder (VAE) simulator and an EfficientNet-B7 classifier architecture embedded within a Siamese twin neural network are presented in this paper in order to demonstrate two use cases. These use cases demonstrated the versatility of our innovative concept in measuring, categorizing, inferring hu-man behavior, and generating gestures for other researchers.

Epilepsy is a prevalent neurological disorder with considerable risks, including physical impairment and irreversible brain damage from seizures. Given these challenges, the urgency for prompt and accurate seizure detection cannot be overstated. Traditionally, experts have relied on manual EEG signal analyses for seizure detection, which is labor-intensive and prone to hu-man error. Recognizing this limitation, the rise of deep learning methods has been heralded as a promising avenue, offering more refined diagnostic precision. On the other hand, the prevailing challenge in many models is their constrained emphasis on specific domains, potentially diminishing their robustness and precision in complex real-world environments. This paper presents a novel model that seamlessly integrates the salient features from the time-frequency domain alongside pivotal statistical attributes derived from EEG signals. This fusion process involves the integration of essential statistics, including the mean, median, and variance, combined with the rich data from compressed time-frequency (CWT) images processed using auto-encoders. This multidimensional feature set provides a robust foundation for subsequent analytic steps. A long short-term memory (LSTM) network, meticulously optimized for the renowned Bonn Epilepsy dataset, was used to enhance the capability of the proposed model. Preliminary evaluations underscore the prowess of the proposed model: remarkable 100% accuracy in most of the binary classifications, impressive performance exceeding 95% for three-class and four-class challenges, and a commendable rate exceeding 93.5% for the five-class classification.

Radiator reliability is crucial in environments characterized by high temperatures and friction, where prompt interventions are often required to prevent system failures. This study introduces a proactive approach to radiator fault diagnosis, leveraging the integration of Gaussian Mixture Model (GMM) and Long-Short Term Memory (LSTM) autoencoders. Vibration signals from radiators were systematically collected through randomized durability vibration bench tests, resulting in four operating states—two normal, one unknown, and one faulty. Time-domain statistical features of these signals were extracted and subjected to Principal Component Analysis (PCA) to facilitate efficient data interpretation. Subsequently, this study discusses the comparative effectiveness of GMM and LSTM in fault detection. GMMs are deployed for initial fault classification, leveraging their clustering capabilities, while LSTM autoencoders excel in capturing time-dependent sequences, facilitating advanced anomaly detection for previously unencountered faults. This alignment offers a potent and adaptable solution for radiator fault diagnosis, particularly in challenging high-temperature or high-friction environments. Consequently, the proposed methodology not only provides a robust framework for early-stage fault diagnosis but also effectively balances diagnostic capabilities during operation. Additionally, this study presents the foundation for advancing reliability life assessment in accelerated life testing, achieved through dynamic threshold adjustments using both the absolute log-likelihood distribution of the GMM and the reconstruction error distribution of the LSTM autoencoder model.

Photovoltaic(PV) power generation is highly nonlinear and stochastic. Accurate prediction of PV power generation plays a crucial role in grid connection as well as the operation and scheduling of power plants. To predict the PV power combination model, this paper suggests a method based on Empirical Mode Decomposition (EMD), Ensemble Empirical Mode Decomposition (EEMD), Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Nonlinear Auto-Regressive Neural Networks with Exogenous Input (NARXNN), Long Short Term Memory (LSTM) Neural Network, and Light Gradient Boosting Machine (LightGBM) algorithms. To attempt to reduce the non-smoothness of PV power, the weather variable features with the greatest effect on PV power are first identified by correlation analysis. Following this, the PV power modal decomposition is split and reorganized into a new feature matrix. Finally， a NARX is used to obtain preliminary PV power components and residual vector features， and the PV power is predicted by combining three models of LightGBM， LSTM, and NARX and then the final prediction results are obtained by combining the PV power prediction results using error inverse method weighted optimization. The prediction results demonstrate that the model put forth in this paper outperforms those of other models and validate the model's validity by utilizing real measurement data from Andre Agassi College in the United States.

Deep neural networks (DNNs) have made a huge impact in the field of machine learning by providing unbeatable humanlike performance to solve real-world problems such as image processing and natural language processing (NLP). Convolutional neural network (CNN) and recurrent neural network (RNN) are two typical architectures that are widely used to solve such problems. Time sequence-dependent problems are generally very challenging, and RNN architectures have made an enormous improvement in a wide range of machine learning problems with sequential input involved. In this paper, different types of RNN architectures are compared. Special focus is put on two well-known gated-RNN’s Long Term Short Memory (LSTM) and Gated Recurrent Unit (GRU). We evaluated these models on the task of force estimation system in pouring. In this study, four different models including multi-layers LSTM, multi-layers GRU, single-layer LSTM and single-layer GRU) were created and trained. The result suggests that multi-layer GRU outperformed other three models.

This paper proposes an interpretable geometry solution based on the formal language set of text and diagram. Geometry problems are solved by machines, which still poses challenges in natural language processing and computer vision. Significant progress promotes existing methods in the extraction of geometric formal languages. However, the neglect of the graph structure information in the formal language and the lack of further refinement of the extracted language set can lead to the poor effect of the theorem prediction and affect the accuracy in problem solving. In this paper, the formal language graph is constructed by the extracted formal language set and used to theorem prediction by graph convolutional network. So as to better extract the relationship set of diagram elements, an improved diagram parser was proposed. The test results indicate that the improved method has good results in solving the problem with interpretability geometry.

Autonomous image recognition has numerous potential applications in the field of planetary science and geology. For instance, having the ability to classify images of rocks would allow geologists to have immediate feedback without having to bring back samples to the laboratory. Also, planetary rovers could classify rocks in remote places and even in other planets without needing human intervention. Shu et al. classified 9 different types of rock images using a Support Vector Machine (SVM) with the image features extracted autonomously. Through this method, the authors achieved a test accuracy of 96.71%. In this research, Convolutional Neural Networks(CNN) have been used to classify the same set of rock images. Results show that a 3-layer network obtains an average accuracy of 99.60% across 10 trials on the test set. A version of Self-taught Learning was also implemented to prove the generalizability of the features extracted by the CNN. Finally, one model has been chosen to be deployed on a mobile device to demonstrate practicality and portability. The deployed model achieves a perfect classification accuracy on the test set, while taking only 0.068 seconds to make a prediction, equivalent to about 14 frames per second.

Objectives: In order to solve the problem that most of the existing research focuses on the binary tactile attributes of objects,which ignores the tactile attribute strength and category recognition,an attribute strength and category recognition method based on convolutional neural network matrix-label is proposed. Methods:Firstly,in the data preparation stage,we preprocess the raw data and determine the matrix labels to build the haptic dataset.Secondly,in the feature extraction stage,we fuse the haptic data of two fingers and use the convolutional neural network to extract the attribute strength features.Finally,in the attribute strength and category recognition stage,all channel haptic data is fused to predict the attribute strength and category.Results:We compared with the multi-label convolutional neural network method in terms of elastic strength,hardness strength and category,and compared the attribute strength recognition capabilities of the two methods using novel objects outside the haptic dataset.The results show that the accuracy of the last 20 iterations of the matrix-label method has an average elastic strength of 96.73%,hardness strength of 97.34%,and category of 96.67%.The performance is better.When the Euclidean distance between the prediction of the novel object and the real label is less than 1,the accuracy of the elastic strength is best to reach 100%,and the hardness strength is best to reach 100%.The performance is better. Conclusions:The effectiveness of the method has been verified.Comparing with the convolutional neural network method,our method can effectively recognize the attribute strength and category of objects.

The detection of seismic events at regional and teleseismic distances is critical to Nuclear Treaty Monitoring. Traditionally, detecting regional and teleseismic events has required the use of an expensive multi-instrument seismic array; however in this work, we present DeepPick, a novel seismic detection algorithm capable of array-like performance from a single trace. We achieve this directly, by training our single-trace detector against labeled events from an array catalog, and by utilizing a deep temporal convolutional neural network. The training data consists of all arrivals in the International Seismological Centre Catalog for seven seismic arrays over a five year window from 1 Jan 2010 to 1 Jan 2015, yielding a total training set of 608,362 detections. The test set consists of the same seven arrays over a one year window from 1 Jan 2015 to 1 Jan 2016. We report our results by training the algorithm on six of the arrays and testing it on the seventh, so as to demonstrate the transportability and generalization of the technique to new stations. Detection performance against this test set is outstanding. Fixing a type-I error rate of 1%, the algorithm achieves an overall recall rate of 73% on the 141,095 array beam picks in the test set, yielding 102,394 correct detections. This is more than 4 times the 23,259 detections found in the analyst-reviewed single-trace catalogs over the same period, and represents an 8dB improvement in detector sensitivity over current methods. These results demonstrate the potential of our algorithm to significantly enhance the effectiveness of the global treaty monitoring network.

In detecting DDoS, deep learning faces challenges and difficulties such as high computational demands, long training times, and complex model interpretation. This research focuses on overcoming these challenges by proposing an effective strategy for detecting DDoS attacks in unbalanced network environments. This research uses SMOTE to increase the class distribution of the data set by allowing models using LSTM to learn time anomalies effectively when DDoS attacks occur. The experiments carried out have shown significant improvement in the performance of the LSTM model when integrated with SMOTE. These include validation loss results of 0.048 for LSTM SMOTE and 0.1943 for LSTM without SMOTE, accuracy of 99.50 and 97.50. Apart from that, there was an increase in the f1 score from 93.4% to 98.3%. In this research, it is proven that SMOTE can be used as an effective strategy to improve model performance in detecting DDoS attacks on heterogeneous networks, as well as increasing model robustness and reliability.

Autonomous mobile robots are essential to the industry, and human-robot interactions are becoming more common nowadays. These interactions require that the robots navigate scenarios with static and dynamic obstacles in a safely manner, avoiding collisions. This paper presents a physical implementation of a method for dynamic obstacle avoidance using a Long Short-Term Memory (LSTM) neural network that obtains information from the mobile robot’s LiDAR for it to be capable of navigating through scenarios with static and dynamic obstacles while avoiding collisions and reaching its goal. The model is implemented using a TurtleBot3 mobile robot within an OptiTrack Motion Capture (MoCap) system for obtaining its position at any given time. The user operates the robot through these scenarios, recording its LiDAR readings, target point, position inside the MoCap system, and its linear and angular velocities, all of which serve as the input for the LSTM network. The model is trained on data from multiple user-operated trajectories across five different scenarios, outputting the linear and angular velocities for the mobile robot. Physical experiments prove that the model is successful in allowing the mobile robot to reach the target point in each scenario while avoiding the dynamic obstacle, with a validation accuracy of 98.02%.

This paper investigates nonverbal communication in human interactions, with a specific focus on facial expressions. Employing a Long Short-Term Memory (LSTM) architecture and a custom-ized facial expression framework, our approach aims to improve virtual agent interactions by incorporating subtle nonverbal cues. The paper contributes to the emerging field of facial expres-sion generation, addressing gaps in current research and presenting a novel framework within Unreal Engine 5. The model's architecture, trained on the CANDOR corpus, captures temporal dynamics, and refines hyperparameters for optimal performance. During testing, the trained model showed a cosine similarity of -0.95. This enables the algorithm to accurately respond to non-verbal cues and interact with humans in a way that is comparable to human-human interac-tion. Unlike other approaches in the field of facial expression generation, the presented method is more comprehensive and enables the integration of a multi-modal approach for generating facial expressions. Future work involves integrating blendshape generation, real-world testing, and the inclusion of additional modalities to create a comprehensive framework for seamless hu-man-agent interactions beyond facial expressions.

In industrial settings, gears play a crucial role by assisting various machinery functions such as speed control, torque manipulation, and altering motion direction.The malfunction or failure of these gear components can have serious repercussions, resulting in production halts and financial losses. As a result, there is an increasing requirement to monitor the state of these components in order to avoid such issues from occurring. To address this need, research efforts have focused on early defect detection in gears in order to reduce the impact of possible failures. This study focused on analyzing vibration and thermal datasets from two extruder machine gearboxes using an autoencoder Long Short-Term Memory (LSTM) model. The major goal is to implement an outlier detection approach to detect and classify defects. The results of this study highlighted the extraordinary performance of the Autoencoder LSTM model, which achieved an impressive accuracy rate of 94.42% in recognizing malfunctioning gearboxes within the extruder machine system. Furthermore, the study used a thorough global metrics evaluation methodology to further test the model’s dependability and efficacy, consequently substantiating the proposed approach’s validity.

Raindrop size distribution (RSD) is an index for reflecting precipitation characteristics. Analyzing the differences of RSD plays an important role in understanding the precipitation microphysical processes and improving quantitative precipitation prediction of radar. In this paper, the RSD data of Dafang (57708) station, Majang (57828) station, and Luodian (57916) station (with an altitude of 1722.7m, 985.0m, and 441.5m, respectively) in Guizhou are analyzed according to the precipitation microphysical characteristics. First,Particles with particle size less than 1mm contributes the highest value to the density of particle number , and decrease with the descending altitude. Second, the GAMMA distribution fit shows a better fit compare to M-P distribution fit, and GAMMA distribution fit increases with the ascending altitude. Third, the mass-weighted average diameter is not sensitive enough to the change of precipitation intensity with correlation coefficients of 54.84%, but there is a obvious relationship between the average volume diameter and the precipitation intensity with correlation coefficients of 69.15%. Then, the precipitation prediction model is established using LSTM neural network after fusing the representative microphysical characteristics of RSD with radar and rain gauge data. The precipitation prediction model is applied to the Dafang (57708) site to predict precipitation for time range of 0-180 minutes, and it is found that for the prediction of convective cloud and stratiform cloud precipitation, the 60-minute prediction results are the most consistent with the actual precipitation, where the correlation coefficients are 92.87% and 92.57%, respectively. Conclusively, the results demonstrate that combining with the RSD base data could improve the reliability of precipitation forecasting.

Activated sludge process is a well known method to treat municipal and industrial waste water. In this complex process, the oxygen concentration in the reactors plays a critical role in the plant efficiency. This paper proposes the use of a Long Short-Term Memory (LSTM) network to identify an input-output model suitable for the design of an oxygen concentration controller. The model is identified from easy-accessible measures collected from a real plant. This dataset covers almost a month. The performances achieved with the proposed LSTM model are compared with the ones obtained with a standard AutoRegressive model with eXogenous input (ARX). Both models catch the oscillation and the overall behaviour (ARX ρ=0.833 , LSTM ρ=0.921), but, while the ARX model fails in reaching the correct amplitude (FIT=41.20%), the LSTM presents satisfactory performance (FIT=60.56%).

This paper realizes early bearing fault warning through bearing fault time series prediction, and proposes a bearing fault time series prediction model based on optimized maximum correlation kurtosis deconvolution (MCKD) and long short-term memory (LSTM) recurrent neural network to ensure bearings operation reliability. The model is based on lifecycle vibration signal of the bearing, to begin, the cuckoo search (CS) is utilized to optimize the parameter filter length L and deconvolution period T of MCKD, taking into account the influence and periodicity of the bearing time series, the fault impact component of the optimized MCKD deconvolution time series is improved. Then select the LSTM learning rate α depending on deconvolution time series. Finally, the dataset obtained through various preprocessing approaches are used to train and predict the LSTM model. The average prediction accuracy of the optimized MCKD-LSTM model is 26 percent higher than that of the original time series, proving the efficiency of this method, and the prediction results track the real fault data well, according to the XI'AN JIAOTONG University XJTU-SY bearing dataset.

In cognitive radio systems, identifying spectrum opportunities is fundamental to efficiently use the spectrum. Spectrum occupancy prediction is a convenient way of revealing opportunities based on previous occupancies. Studies have demonstrated that usage of the spectrum has a high correlation over multidimensions which includes time, frequency, and space. Accordingly, recent literature uses tensor-based methods to exploit the multidimensional spectrum correlation. However, these methods share two main drawbacks. First, they are computationally complex. Second, they need to re-train the overall model when no information is received from any base station for any reason. Different than the existing works, this paper proposes a method for dividing the multidimensional correlation exploitation problem into a set of smaller sub-problems. This division is achieved through composite two-dimensional (2D)-long short-term memory (LSTM) models. Extensive experimental results reveal a high detection performance with more robustness and less complexity attained by the proposed method. The real-world measurements provided by one of the leading mobile network operators in Turkey validate these results.

Several researchers have contemplated deep learning-based post-filters to increase the quality of statistical parametric speech synthesis, which perform a mapping of the synthetic speech to the natural speech, considering the different parameters separately and trying to reduce the gap between them. The Long Short-term Memory (LSTM) Neural Networks have been applied successfully in this purpose, but there are still many aspects to improve in the results and in the process itself. In this paper, we introduce a new pre-training approach for the LSTM, with the objective of enhancing the quality of the synthesized speech, particularly in the spectrum, in a more efficient manner. Our approach begins with an auto-associative training of one LSTM network, which is used as an initialization for the post-filters. We show the advantages of this initialization for the enhancing of the Mel-Frequency Cepstral parameters of synthetic speech. Results show that the initialization succeeds in achieving better results in enhancing the statistical parametric speech spectrum in most cases when compared to the common random initialization approach of the networks.

This study develops a 7-layer Long Short-Term Memory (LSTM) model to enhance early diabetes detection in Oman, aligning with the theme of 'Artificial Intelligence in Healthcare'. The model focuses on addressing the increasing prevalence of Type 2 diabetes, projected to impact 23.8% of Oman's population by 2050. It employs LSTM neural networks to manage factors contributing to this rise, including obesity and genetic predispositions, and aims to bridge the gap in public health awareness and prevention. The model's performance is evaluated through various metrics. It achieves an accuracy of 99.40%, specificity and sensitivity of 100% for positive cases, a recall of 99.34% for negative cases, an F1 score of 96.24%, and an AUC score of 94.51%. These metrics indicate the model's capability in diabetes detection. The implementation of this LSTM model in Oman's healthcare system is proposed to enhance early detection and prevention of diabetes. This approach reflects an application of AI in addressing a significant health concern, with potential implications for similar healthcare challenges globally.st diagnostic capabilities, representing a significant leap forward in healthcare technology in Oman.

Peanut leaf spot is a worldwide disease whose prevalence poses a major threat to peanut yield and quality, and accurate prediction models are urgently needed for timely disease management. In this study, we proposed a novel peanut leaf spot prediction method based on an improved long short-term memory (LSTM) model and multi-year meteorological data combined with disease survey records. Our method employed a combination of convolutional neural network (CNN) and LSTM to capture spatial-temporal patterns from the data and improve the model ability to recognize dynamic features of the disease. In addition, we introduced an attention mechanism module to enhance model performance by focusing on key features. Through several hyper-parameter optimization adjustments, we identified a peanut leaf spot disease condition index prediction model with a learning rate of 0.001, a number of cycles (Epochs) of 800, and an optimizer of Adma. The results showed that the integrated model demonstrated excellent prediction ability, obtaining an RMSE of 0.063 and an R2 of 0.951, which reduced the RMSE by 0.253 and 0.204, and raised the R2 by 0.155 and 0.122, respectively, compared to the single CNN and LSTM. Predicting the incidence and severity of peanut leaf spot disease based on the meteorological conditions and neural networks is feasible and valuable to help growers made accurate management decisions and reduced disease impacts through optimal fungicide application timing.

The performance of bearings plays a pivotal role in determining the dependability and security of rotating machinery. In intricate systems demanding exceptional reliability and safety, the ability to accurately forecast fault occurrences during operation holds profound significance. Such predictions serve as invaluable guides for crafting well-considered reliability strategies and executing maintenance practices aimed at enhancing reliability. In order to ensure the reliability of bearing operation, this article investigates the application of three advanced techniques—Maximum Correlation Kurtosis Deconvolution (MCKD), Multi-Scale Permutation Entropy (MPE), and Long Short-Term Memory (LSTM) recurrent neural networks—for the prediction of the remaining useful life (RUL) of rolling bearings. Each technique's principles, methodologies, and applications are comprehensively reviewed, offering insights into their respective strengths and limitations. Case studies and experimental evaluations are presented to assess their performance in RUL prediction. Findings reveal that MCKD enhances fault signatures, MPE captures complexity, and LSTM excels in modeling temporal patterns. The root mean square error of the prediction results is 0.007. The fusion of these techniques offers a comprehensive approach to RUL prediction, leveraging their unique attributes for more accurate and reliable predictions.

Celiac disease is a gluten-sensitive immune-mediated enteropathy of the small intestine that occurs in genetically predisposed individuals. Abnormal immune response results in mucosal inflammation, villous atrophy, and crypt hyperplasia. This study was a proof-of-concept exercise that used a convolutional neural network to classify hematoxylin and eosin (H&E) histological images of celiac disease, normal small intestine control, and non-specified duodenal inflammation; 7294, 11,642, and 5966 images, respectively. The trained network classified celiac disease images with high performance (accuracy 99.7%, precision 99.6%, recall 99.3%, F1-score 99.5%, and specificity 99.8%). Interestingly, when the same network (already trained for the 3 class images), analyzed duodenal adenocarcinoma (3723 images), the new images were classified as duodenal inflammation in 63.65%, small intestine control in 34.73%, and celiac disease in 1.61% of the cases. Finally, when the network was retrained using the 4 histological subtypes of images, all performance parameters were above 99% for celiac disease. In conclusion, the convolutional neural network (CNN)-based deep neural system was able to classify medical histological images with high performance. Narrow artificial intelligence (AI) is designed to perform tasks that typically require human intelligence, but it operates within limited constraints and is task specific.