Search Results (130)

In recent years, the proliferation of forged videos, known as deepfakes, has escalated significantly, primarily due to advancements in technologies such as Generative Adversarial Networks (GANs), diffusion models, and Vision Language Models (VLMs). These deepfakes present substantial risks, threatening political stability, facilitating celebrity impersonation, and enabling tampering with evidence. As the sophistication of deepfake technology increases, detecting these manipulated videos becomes increasingly challenging. Most of the existing deepfake detection methods use Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), or Vision Transformers (ViTs), achieving strong accuracy but exhibiting high computational demands. This highlights the need for a lightweight yet effective pipeline for real-time and resource-limited scenarios. This study introduces a lightweight deep learning model for deepfake detection in order to address this emerging threat. The model incorporates three integral components: MobileNet for feature extraction, a Convolutional Block Attention Module (CBAM) for feature enhancement, and a Gated Recurrent Unit (GRU) for temporal analysis. Additionally, a pre-trained Multi-Task Cascaded Convolutional Network (MTCNN) is utilized for face detection and cropping. The model is evaluated using the Deepfake Detection Challenge (DFDC) and Celeb-DF v2 datasets, demonstrating impressive performance, with 98.2% accuracy and a 99.0% F1-score on Celeb-DF v2 and 95.0% accuracy and a 97.2% F1-score on DFDC, achieving a commendable balance between simplicity and effectiveness. Full article

This work introduces a unified Artificial Intelligence-based framework for the optimal tuning of gains in a neural discrete-time sliding mode controller (SMC) applied to a two-degree-of-freedom robotic manipulator. The novelty lies in combining surrogate-assisted optimization with normalized search spaces to enable a fair comparative analysis of three metaheuristic strategies: Bayesian Optimization (BO), Particle Swarm Optimization (PSO), and Genetic Algorithms (GAs). The manipulator dynamics are identified via a discrete-time recurrent high-order neural network (NN) trained online using an Extended Kalman Filter with adaptive noise covariance updates, allowing the model to accurately capture unmodeled dynamics, nonlinearities, parametric variations, and process/measurement noise. This neural representation serves as the predictive plant for the discrete-time SMC, enabling precise control of joint angular positions under sinusoidal phase-shifted references. To construct the optimization dataset, MATLAB^® simulations sweep the controller gains $(k_0^{*},k_1^{*})$ over a bounded physical domain, logging steady-state tracking errors. These are normalized to mitigate scaling effects and improve convergence stability. Optimization is executed in Python^® using integrated scikit-learn, DEAP, and scikit-optimize routines. Simulation results reveal that all three algorithms reach high-performance gain configurations. Here, the combined cost is the normalized aggregate objective $\tilde{J}$ constructed from the steady-state tracking errors of both joints. Under identical experimental conditions (shared data loading/normalization and a single Python pipeline), PSO attains the lowest error in Joint 1 ($7.36\times {10}^{-5}$ rad) with the shortest runtime (23.44 s); GA yields the lowest error in Joint 2 ($8.18\times {10}^{-3}$ rad) at higher computational expense (≈69.7 s including refinement); and BO is competitive in both joints ($7.81\times {10}^{-5}$ rad, $8.39\times {10}^{-3}$ rad) with a runtime comparable to PSO (23.65 s) while using only 50 evaluations. Full article

In order to address the challenge of achieving optimal prediction accuracy when a single prediction model faced with changes in meteorological conditions of different weather types, this paper proposes a photovoltaic (PV) power prediction method based on the combination of similar day clustering and convolutional neural network (CNN)-gated recurrent unit (GRU). The Pearson correlation coefficient and Spearman’s correlation coefficient are used to filter out the key features such as total solar radiation and module temperature to construct a new input dataset; the K-means algorithm is used to perform clustering analysis on the data, and the data are classified into sunny, cloudy, and rainy days; the spatial correlation features of the meteorological factors are extracted by using the convolutional neural network (CNN), and the CNN-GRU model is established by combining with the gated recurrent units (GRUs). The PV output power is predicted based on the PV power data and the corresponding meteorological data from a place in Ningxia, collected during June to August 2020, and the method proposed in the article is tested. Validation results show that, compared to other models, the model proposed in this paper reduces MAE and RMSE by 66.1% and 65.7% on average under three different weather type scenarios, and improves R² by 19.8% on average. This verifies that the model has high prediction accuracy and generalization ability, achieving better results in PV output power prediction. The CNN-GRU model demonstrates superior capability in modeling short- and long-term dependencies compared to other deep learning hybrid approaches, while also achieving higher computational efficiency and faster training convergence. Full article

Sound event localization and detection (SELD) is a fundamental task in spatial audio processing that involves identifying both the type and location of sound events in acoustic scenes. Current SELD models often struggle with low signal-to-noise ratios (SNRs) and high reverberation. This article addresses SELD by reformulating direction of arrival (DOA) estimation as a multi-class classification task, leveraging deep convolutional recurrent neural networks (CRNNs). We propose and evaluate two modified architectures: M-DOAnet, an optimized version of DOAnet for localization and tracking, and M-SELDnet, a modified version of SELDnet, which has been designed for joint SELD. Both modified models were rigorously evaluated on the STARSS23 dataset, which comprises 13-class, real-world indoor scenes totaling over 7 h of audio, using spectrograms and acoustic intensity maps from first-order Ambisonics (FOA) signals. M-DOAnet achieved exceptional localization (6.00° DOA error, 72.8% F1-score) and perfect tracking (100% MOTA with zero identity switches). It also demonstrated high computational efficiency, training in 4.5 h (164 s/epoch). In contrast, M-SELDnet delivered strong overall SELD performance (0.32 rad DOA error, 0.75 F1-score, 0.38 error rate, 0.20 SELD score), but with significantly higher resource demands, training in 45 h (1620 s/epoch). Our findings underscore a clear trade-off between model specialization and multifunctionality, providing practical insights for designing SELD systems in real-time and computationally constrained environments. Full article

Modelling events that change over time is one of the most difficult problems in data analysis. Forecasting of time-varying electric power values is also an important problem in data analysis. Regression methods, machine learning, and deep learning methods are used to learn different patterns from data and develop a consumption prediction model. The aim of this study is to determine the most successful models for short-term power consumption prediction with deep learning and to achieve the highest prediction accuracy. In this study, firstly, the data was evaluated and organized with exploratory data analysis (EDA) on a ready dataset and the features of the data were extracted. Studies were carried out on long short-term memory (LSTM), gated recurrent unit (GRU), simple recurrent neural networks (SimpleRNN) and bidirectional long short-term memory (BiLSTM) architectures. First, four architectures were used with 11 different optimization methods. In this study, it was seen that a high success rate of 0.9972 was achieved according to the R² score index. In the following, the first study was tried with different epoch numbers. Afterwards, this study was carried out with 264 separate models produced using four architectures, 11 optimization methods, and six activation functions in order. The results of all these studies were obtained according to the root mean square error (RMSE), mean absolute error (MAE), and R²_score indexes. The R²_score indexes graphs are presented. Finally, the 10 most successful applications are listed. Full article

Background and Objectives: Gliomas are characterized by high disability rates, frequent recurrence, and low survival rates, posing a significant threat to human health. Accurate grading of gliomas is crucial for treatment plan selection and prognostic assessment. Previous studies have primarily focused on the binary classification (i.e., high grade vs. low grade) of gliomas. In order to perform the four-grade (grades I, II, III, and IV) glioma classification preoperatively, we constructed an artificial neural network (ANN) model using magnetic resonance imaging data. Materials and Methods: We reviewed and included patients with gliomas who underwent preoperative MRI examinations. Radiomics features were derived from contrast-enhanced T1-weighted images (CE-T₁WI) using Pyradiomics and were selected based on their Spearman’s rank correlation with glioma grades. We developed an ANN model to classify the four pathological grades of glioma, assigning training and validation sets at a 3:1 ratio. A diagnostic confusion matrix was employed to demonstrate the model’s diagnostic performance intuitively. Results: Among the 362-patient cohort, the ANN model’s diagnostic performance plateaued after incorporating the first 19 of the 530 extracted radiomic features. At this point, the average overall diagnostic accuracy ratings for the training and validation sets were 91.28% and 87.04%, respectively, with corresponding coefficients of variation (CVs) of 0.0190 and 0.0272. The diagnostic accuracies for grades I, II, III, and IV in the training set were 91.9%, 89.9%, 92.1%, and 90.7%, respectively. The diagnostic accuracies for grades I, II, III, and IV in the validation set were 88.7%, 87.1%, 86.5%, and 86.9%, respectively. Conclusions: The MRI radiomics-based ANN model shows promising potential for the four-type classification of glioma grading, offering an objective and noninvasive method for more refined glioma grading. This model could aid in clinical decision making regarding the treatment of patients with various grades of gliomas. Full article

To improve the accuracy of leak identification and location of water supply pipelines, a novel convolution gated recurrent unit method based on the attention mechanism is proposed in this paper. Firstly, a convolutional neural network is used to capture the localspatio-temporal characteristics of the signal. Secondly, a gated recurrent unit is used to extract the signal’s long dependence relationship. Finally, an attention mechanism is combined to highlight the influence of key features in the learning process, so as to achieve accurate recognition of the pipeline pressure state. The accurate identification of leakage faults is expected to further improve the location accuracy of pipeline leakage points, which is very important for the practical application of the algorithm in engineering. In order to verify the effectiveness of the proposed method, a simulated leakage test platform is set up for the leakage simulation test. The test results of different leakage conditions show that the recognition accuracy of the proposed network structure is 98.75% for test samples, which is higher than other network structures of the same type. According to the identification results of leakage characteristics, the VMD method is used to extract the high-frequency components of the negative pressure wave signal, so as to obtain the inflection point of the negative pressure wave, so as to determine the arrival time difference of the signal, and the arrival time method based on the negative pressure wave is used to locate the leakage point. Across 12 leak locations, the maximum relative error is 7.67%, the minimum relative error is 0.86%, and the average relative error is only 2.97%, achieving the best performance among the various methods. The positioning accuracy meets the requirement of practical application and the algorithm has good robustness. Full article

To handle and manage battery degradation in electric vehicles (EVs), various capacity estimation methods have been proposed and can mainly be divided into traditional modeling methods and data-driven methods. For realistic conditions, data-driven methods take the advantage of simple application. However, state-of-the-art machine learning (ML) algorithms are still kinds of black-box models; thus, the algorithms do not have a strong ability to describe the inner reactions or degradation information of batteries. Due to a lack of interpretability, machine learning may not learn the degradation principle correctly and may need to depend on big data quality. In this paper, we propose a physics-informed recurrent neural network (PIRNN) with a fractional-order gradient for fast battery degradation estimation in running EVs to provide a physics-informed neural network that can make algorithms learn battery degradation mechanisms. Incremental capacity analysis (ICA) was conducted to extract aging characteristics, which could be selected as the inputs of the algorithm. The fractional-order gradient descent (FOGD) method was also applied to improve the training convergence and embedding of battery information during backpropagation; then, the recurrent neural network was selected as the main body of the algorithm. A battery dataset with fast degradation from ten EVs with a total of 5697 charging snippets were constructed to validate the performance of the proposed algorithm. Experimental results show that the proposed PIRNN with ICA and the FOGD method could control the relative error within $5\%$ for most snippets of the ten EVs. The algorithm could even achieve a stable estimation accuracy (relative error < 3%) during three-quarters of a battery’s lifetime, while for a battery with dramatic degradation, it was difficult to maintain such high accuracy during the whole battery lifetime. Full article

The electrical submersible pump (ESP) well system is widely used in the oil industry due to its advantages of high displacement and lift capability. However, it is associated with significant energy consumption. In order to conserve electrical energy and enhance the efficiency of petroleum companies, a deep learning-based energy consumption calculation method is proposed and utilized to optimize the most energy-efficient operating regime. The energy consumption of the ESP well system is precisely determined through the application of the Pearson correlation coefficient analysis method, which is utilized to examine the relationship between production parameters and energy usage. This process aids in identifying the input parameters of the model. Following this, an energy consumption prediction model is developed using the dual-stage attention-based recurrent neural network (DA-RNN) algorithm. To evaluate the accuracy of the DA-RNN model, a comparison of its errors is carried out in comparison to three other deep learning algorithms: Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and Transform. Lastly, an orthogonal experiment is executed using the chosen model to pinpoint the most energy-efficient operating regime. Analysis of 325 ESP wells in the Bohai PL oil field indicated that ten parameters, including choke diameter, casing pressure, pump inlet pressure, pump outlet pressure, motor temperature, frequency, oil production, gas production, water production, and GOR significantly impact the energy consumption of the ESP well system. Consequently, these parameters were selected as input variables for the deep learning model. Due to the attention mechanisms employed in the encoding and decoding stages, the DA-RNN algorithm achieved the best performance during model evaluation and was chosen for constructing the energy consumption prediction model. Furthermore, the DA-RNN algorithm demonstrates better model generalization capabilities compared to the other three algorithms. Based on the energy consumption prediction model, the operating regime of the ESP system was optimized to save up to 12% of the maximum energy. The energy consumption of the ESP well system is affected by numerous parameters, and it is difficult to comprehensively evaluate and predict quantitatively. Thus, this work proposes a data-driven model based on the DA-RNN algorithm, which has a dual-stage attention mechanism to rapidly and accurately predict the energy consumption of the ESP well system. Optimization of production parameters using this model can effectively reduce energy consumption. Full article

This work proposes and describes a deep learning-based approach utilizing recurrent neural networks (RNNs) for beam pattern synthesis considering uniform linear arrays. In this particular case, the deep neural network (DNN) learns from previously optimized radiation patterns as inputs and generates complex excitations as output. Beam patterns are optimized using a genetic algorithm during the training phase in order to reduce sidelobes and achieve high directivity. Idealized and test beam patterns are employed as inputs for the DNN, demonstrating their effectiveness in scenarios with high prediction complexity and closely spaced elements. Additionally, a comparative analysis is conducted among the three DNN architectures. Numerical experiments reveal improvements in performance when using the long short-term memory network (LSTM) compared to fully connected and convolutional neural networks. Full article

Evaluating students’ learning effectiveness is of great importance for gaining a deeper understanding of the learning process, accurately diagnosing learning barriers, and developing effective teaching strategies. Emotion, as a key factor influencing learning outcomes, provides a novel perspective for identifying cognitive states and emotional experiences. However, traditional evaluation methods suffer from one sidedness in feature extraction and high complexity in model construction, often making it difficult to fully explore the deep value of emotional data. To address this challenge, we have innovatively proposed a lightweight neurodynamic model: 3D-BCLAM. This model cleverly integrates Bidirectional Convolutional Long Short-Term Memory (BCL) and dynamic attention mechanism, in order to efficiently capture emotional dynamic changes in time series with extremely low computational cost. 3D-BCLAM can achieve a comprehensive evaluation of students’ learning outcomes, covering not only the cognitive level but also delving into the emotional dimension for detailed analysis. Under testing on public datasets, 3D-BCLAM has demonstrated outstanding performance, significantly outperforming traditional machine learning and deep learning models based on Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). This achievement not only validates the effectiveness of the 3D-BCLAM model, but also provides strong support for promoting the innovation of student learning effectiveness assessment. Full article

This paper presents an online model-free sensor fault-tolerant control scheme capable of tolerating the most common faults affecting an induction motor. This approach involves using neural networks for fault detection to provide the controller with sufficient information to counteract adverse consequences due to sensor faults, such as degradation in performance, reliability, and even failures in the control system. The proposed approach does not consider the knowledge of the nominal model of the system or when the fault may occur. Therefore, a high-order recurrent neural network trained online by the Extended Kalman Filter is used to obtain a mathematical model of the system. The obtained model is used to synthesize a discrete-time sliding mode control. Then, the fault-detection and -isolation stage is performed by independent neural networks, which have as input the signal from the current sensor and the position sensor, respectively. In this way, the neural classifiers continuously monitor the sensors, showing the ability to know the sensor status. The combination of controller and fault detection maintains the operation of the motor during the time of the fault occurrence, whether due to sensor disconnection, degradation, or connection failure. In fact, the MLP neural network achieves an accuracy between 95% and 99% and shows an AUC of 97% to 99%, and this neural network correctly classifies true positives with acceptable performance. The Recall value is high, between 97% and 99%, and the F1 score confirms a good performance. In contrast, the CNN shows a higher accuracy, between 96% and 99% in accuracy and 98% to 99% in AUC. In addition, its Recall and F1 reflect a better balance and capacity to handle complex data, demonstrating its superiority to MLP in fault classification. Therefore, neural networks are a promising approach in areas such as fault-tolerant control. Full article

Deep learning models, such as recurrent neural network (RNN) models, are suitable for modeling and forecasting non-stationary time series but are not interpretable. A prediction model with interpretability and high accuracy can improve decision makers’ trust in the model and provide a basis for decision making. This paper proposes a double decomposition strategy based on wavelet decomposition (WD) and empirical mode decomposition (EMD). We construct a prediction model of high-order fuzzy cognitive maps (HFCM), called the WE-HFCM model, which considers interpretability and strong reasoning ability. Specifically, we use the WD and EDM algorithms to decompose the time sequence signal and realize the depth extraction of the signal’s high-frequency, low-frequency, time-domain, and frequency domain features. Then, the ridge regression algorithm is used to learn the HFCM weight vector to achieve modeling prediction. Finally, we apply the proposed WE-HFCM model to stationary and non-stationary datasets in simulation experiments. We compare the predicted results with the autoregressive integrated moving average (ARIMA) and long short-term memory (LSTM) models.For stationary time series, the prediction accuracy of the WE-HFCM model is about 45% higher than that of the ARIMA, about 35% higher than that of the SARIMA model, and about 16% higher than that of the LSTM model. For non-stationary time series, the prediction accuracy of the WE-HFCM model is 69% higher than that of the ARIMA and SARIMA models. Full article

This paper presents two intelligent robust control strategies applied to manage the dynamics of a DC-DC bidirectional buck–boost converter, which is used in conjunction with a supercapacitor as an auxiliary energy system (AES) for regenerative braking in electric vehicles. The Neural Inverse Optimal Controller (NIOC) and the Neural Sliding Mode Controller (NSMC) utilize identifiers based on Recurrent High-Order Neural Networks (RHONNs) trained with the Extended Kalman Filter (EKF) to track voltage and current references from the converter circuit. Additionally, a driving cycle test tailored specifically for typical urban driving in electric vehicles (EVs) is implemented to validate the efficacy of the proposed controller and energy improvement strategy. The proposed NSMC and NIOC are compared with a PI controller; furthermore, an induction motor and its corresponding three-phase inverter are incorporated into the EV control scheme which is implemented in Matlab/Simulink using the “Simscape Electrical” toolbox. The Mean Squared Error (MSE) is computed to validate the performance of the neural controllers. Additionally, the improvement in the State of Charge (SOC) for an electric vehicle battery through the control of buck–boost converter dynamics is addressed. Finally, several robustness tests against parameter changes in the converter are conducted, along with their corresponding performance indices. Full article

The integration of vehicle-to-grid (V2G) technology into smart energy management systems represents a significant advancement in the field of energy suppliers for Industry 4.0. V2G systems enable a bidirectional flow of energy between electric vehicles and the power grid and can provide ancillary services to the grid, such as peak shaving, load balancing, and emergency power supply during power outages, grid faults, or periods of high demand. In this context, reliable prediction of the availability of V2G as an energy source in the grid is fundamental in order to optimize both grid stability and economic returns. This requires both an accurate modeling framework that includes the integration and pre-processing of readily accessible data and a prediction phase over different time horizons for the provision of different time-scale ancillary services. In this research, we propose and compare two data-driven predictive modeling approaches to demonstrate their suitability for dealing with quasi-periodic time series, including those dealing with mobility data, meteorological and calendrical information, and renewable energy generation. These approaches utilize publicly available vehicle tracking data within the floating car data paradigm, information about meteorological conditions, and fuzzy weekend and holiday information to predict the available aggregate capacity with high precision over different time horizons. Two data-driven predictive modeling approaches are then applied to the selected data, and the performance is compared. The first approach is Hankel dynamic mode decomposition with control (HDMDc), a linear state-space representation technique, and the second is long short-term memory (LSTM), a deep learning method based on recurrent nonlinear neural networks. In particular, HDMDc performs well on predictions up to a time horizon of 4 h, demonstrating its effectiveness in capturing global dynamics over an entire year of data, including weekends, holidays, and different meteorological conditions. This capability, along with its state-space representation, enables the extraction of relationships among exogenous inputs and target variables. Consequently, HDMDc is applicable to V2G integration in complex environments such as smart grids, which include various energy suppliers, renewable energy sources, buildings, and mobility data. Full article