Deep learning-based medical image analysis has shown strong potential in disease categorization, segmentation, detection, and even prediction. However, in high-stakes and complex domains like healthcare, the opaque nature of these models makes it challenging to trust predictions, particularly in uncertain cases. This sort of uncertainty can be crucial in medical image analysis; diabetic retinopathy is an example where even slight errors without an indication of confidence can have adverse impacts. Traditional deep learning models rely on single-point predictions, limiting their ability to provide uncertainty measures essential for robust clinical decision-making. To solve this issue, Bayesian approximation approaches have evolved and are gaining market traction. In this work, we implemented a transfer learning approach, building upon the DenseNet-121 convolutional neural network to detect diabetic retinopathy, followed by Bayesian extensions to the trained model. Bayesian approximation techniques, including Monte Carlo Dropout, Mean Field Variational Inference, and Deterministic Inference, were applied to represent the posterior predictive distribution, allowing us to evaluate uncertainty in model predictions. Our experiments on a combined dataset (APTOS 2019 + DDR) with pre-processed images showed that the Bayesian-augmented DenseNet-121 outperforms state-of-the-art models in test accuracy, achieving 97.68% for the Monte Carlo Dropout model, 94.23% for Mean Field Variational Inference, and 91.44% for the Deterministic model. We also measure how certain the predictions are, using an entropy and a standard deviation metric for each approach. We also evaluated the model using both AUC and accuracy scores at multiple data retention levels. In addition to overall performance boosts, these results highlight that Bayesian deep learning does not only improve classification accuracy in the detection of diabetic retinopathy but also reveals beneficial insights about how uncertainty estimation can help build more trustworthy clinical decision-making solutions.

This paper addresses the challenging problem of Unit Commitment (UC), which involves the optimal scheduling of power generation units while adhering to numerous network operational constraints called security-constrained UC (SCUC). SCUC problem aims to minimize costs subject to turning on economically efficient generators and turning off expensive ones. These operational constraints include load balancing, voltage level at buses, minimum up and down time requirements, spinning reserve, and ramp up and down constraints. The SCUC problem, subject to these operational constraints, is a complex mixed-integer nonlinear problem (MINLP). There has been a growing interest in using evolutionary algorithms (EAs) to tackle large-scale multi-objective MINLP problems in recent two decades. This paper introduces a novel approach to address the SCUC problem, which is further complicated by including network constraints. They are pioneering the integration of single and multi-objective EAs to solve the SCUC problem while incorporating AC network constraints through hybrid binary and real coded operators. The development of an ensemble algorithm that combines mixed real and binary coded operators, extended by a bidirectional coevolutionary algorithm to tackle multi-objective SCUC problems. The paper implements a new formulation based on three conflicting objective functions: cost of energy supplied, startup and shutdown costs of generators, energy loss, and voltage deviation to solve the SCUC problem. Implementing a new formulation also addresses the solution of single and multi-objective SCUC problems using a combination of proposed technical and economic objective functions. The proposed algorithm is rigorously tested on a 10-unit IEEE RTS system and a 6-unit IEEE 30-bus test system, both with and without security constraints, addressing week-ahead and day-ahead SCUC scenarios. Simulation results show that the proposed algorithm finds near-global optimal solutions compared to other state-of-the-art EAs. Additionally, the research demonstrates the effectiveness of the proposed search operator by integrating it with a multi-objective coevolutionary algorithm driven by both feasible and infeasible solutions, showcasing superior performance in solving multi-objective SCUC problems. These results are compared with various recently implemented Multi-Objective Evolutionary Algorithms (MOEAs), demonstrating the superiority of

Secondary frequency control systems, such as Automatic Generation Control (AGC), are used in interconnected power grids. However, when a system failure causes systems to separate into zones (islands), AGC can no longer be used, and the primary frequency control is the only control available. Moreover, load changes may cause frequency drop in some areas and over-frequency in other areas. Therefore, the goal in this article will be to design a neural network-based proportional, integral, and derivative (PID) controller in the primary control architecture to control the over frequency condition. The proposed controller is adaptively optimized in two stages by the honey badger algorithm (HBA). In the first stage, the PID controller gain values are optimized by the HBA algorithm for different values of load loss. While in the second stage, a feed-forward artificial neural network (ANN) is trained to match the tie-line measured power to the corresponding optimized HBA-PID gains obtained in the first stage. Finally, the proposed controller is implemented on a two-area interconnected thermal power system. The proposed controller results qualitatively outperform one of the best tuning methods, the Ziegler-Nichols (ZN) approach, and they show that the proposed controller has better dynamic responses with minimal frequency deviations and fast settling time, creating and guaranteeing a margin of stability for the closed loop.

The rapid advancement of computer technology and the widespread adoption of online social media platforms have inadvertently provided fertile ground for individuals with antisocial inclinations to thrive, ushering in a range of security concerns, including the proliferation of fake profiles, hate speech, social bots, and the spread of unfounded rumors. Among these issues, a prominent concern is the prevalence of hate speech within online social networks (OSNs). However, the relevance of numerous studies on hate speech detection has been limited, as they primarily focus on a single language, often English. In response, our research embarks on an exhaustive exploration of multilingual hate speech across 12 distinct languages, offering a novel approach by adapting hate speech detection resources across linguistic boundaries. This study presents the development of a robust, lightweight and multilingual hate speech detection model, known as MLHS-CGCapNet, which combines convolutional and bidirectional gated recurrent units with a capsule network.With commendable accuracy, recall and f-score values of 0.89, 0.80, and 0.84, respectively, our proposed model exhibits strong performance, even when handling an imbalanced dataset. Notably, during the training and validation phases, the suggested model showcases exceptional effectiveness, achieving accuracy values of 0.93 and 0.90, respectively, particularly in the challenging context of imbalanced data. In comparison to both baseline and state-of-the-art techniques, our model offers superior performance.

The classification of vehicles presents notable challenges within the domain of image processing. Traditional models suffer from inefficiency, prolonged training times for datasets, intricate feature extraction, and variable assignment complexities for classification. Conventional methods applied to categorize vehicles from extensive datasets often lead to errors, misclassifications, and unproductive outcomes. Consequently, leveraging machine learning techniques emerges as a promising solution to tackle these challenges. This study adopts a machine learning approach to alleviate image misclassifications and manage large quantities of vehicle images effectively. Specifically, a contrast enhancement technique is employed in the pre-processing stage to highlight pixel values in vehicle images. In the feature segmentation stage, Mask-R-CNN is utilized to categorize pixels into predefined classes. VGG16 is then employed to extract features from vehicle images, while an autoencoder aids in selecting features by learning non-linear input features and compressing representation features. Finally, the CatBoost (CB) algorithm is implemented for vehicle classification (VC) in diverse critical environments, such as inclement weather, twilight, and instances of vehicle blockage. Extensive experiments are conducted using different large-scale datasets with various machine learning platforms. The findings indicate that CB (presumably a specific method or algorithm) attains the highest level of performance on the large-scale dataset named UFPR-ALPR, with an accuracy rate of 98.89%.

This study presents an augmented hybrid approach for improving the diagnosis of malignant skin lesions by combining convolutional neural network (CNN) predictions with selective human interventions based on prediction confidence. The algorithm retains high-confidence CNN predictions while replacing low-confidence outputs with expert human assessments to enhance diagnostic accuracy. A CNN model utilizing the EfficientNetB3 backbone is trained on datasets from the ISIC-2019 and ISIC-2020 SIIM-ISIC melanoma classification challenges and evaluated on a 150-image test set. The model’s predictions are compared against assessments from 69 experienced medical professionals. Performance is assessed using receiver operating characteristic (ROC) curves and area under curve (AUC) metrics, alongside an analysis of human resource costs. The baseline CNN achieves an AUC of 0.822, slightly below the performance of human experts. However, the augmented hybrid approach improves the true positive rate to 0.782 and reduces the false positive rate to 0.182, delivering better diagnostic performance with minimal human involvement. This approach offers a scalable, resource-efficient solution to address variability in medical image analysis, effectively harnessing the complementary strengths of expert humans and CNNs.