Recently, the Ethereum smart contracts have seen a surge in interest from the scientific community and new commercial uses. However, as online trade expands, other fraudulent practices—including phishing, bribery, and money laundering—emerge as significant challenges to trade security. This study is useful for reliably detecting fraudulent transactions; this work developed a deep learning model using a unique metaheuristic optimization strategy. The new optimization method to overcome the challenges, Optimized Genetic Algorithm-Cuckoo Search (GA-CS), is combined with deep learning. In this research, a Genetic Algorithm (GA) is used in the phase of exploration in the Cuckoo Search (CS) technique to address a deficiency in CS. A comprehensive experiment was conducted to appraise the efficiency and performance of the suggested strategies compared with those of various popular techniques, such as k-nearest neighbors (KNN), logistic regression (LR), multi-layer perceptron (MLP), XGBoost, light gradient boosting machine (LGBM), random forest (RF), and support vector classification (SVC), in terms of restricted features and we compared their performance and efficiency metrics to the suggested approach in detecting fraudulent behavior on Ethereum. The suggested technique and SVC models outperform the rest of the models, with the highest accuracy, while deep learning with the proposed optimization strategy outperforms the RF model, with slightly higher performance of 99.71% versus 98.33%.

Depression is a common illness worldwide with doubtless severe implications. Due to the absence of early identification and treatment for depression, millions of individuals worldwide suffer from mental illnesses. It might be difficult to identify those who are experiencing mental health illnesses and to provide them with the early help that they need. Additionally, depression may be associated with thoughts of suicide. Currently, there are no clinically specific diagnostic biomarkers that can identify the severity and type of depression. In this research paper, the novel particle swarm-cuckoo search (PS-CS) optimization algorithm is proposed instead of the traditional backpropagation algorithm for training deep neural networks. The backpropagation algorithm is widely used for supervised learning in deep neural networks, but it has limitations in terms of convergence speed and the possibility of getting trapped in local optima. These problems were addressed by using a deep neural network architecture for depression detection tasks along with the PS-CS optimization technique. The PS-CS algorithm combines the strengths of both particle swarm optimization and cuckoo search algorithms, which allows for a more efficient and effective optimization of the network parameters. We also evaluated how well the suggested methods performed against the most widely used classification models, including (K-nearest neighbor) KNN, (support vector regression) SVR, and decision trees, as well as the most widely used deep learning models, including residual neural network (ResNet), visual geometry group (VGG), and simple neural network (LeNet). The findings show that the suggested method, PS-CS, in conjunction with the CNN model, outperformed all other models, achieving the maximum accuracy of 99.5%. Other models, such as the KNN, decision trees, and logistic regression, achieved lower accuracies ranging from 69% to 97%.

Hyperspectral anomaly detection (HAD) is vital in remote sensing and surveillance, where the goal is to identify rare or unknown targets without prior labels. However, HAD remains challenging due to: (1) high spectral dimensionality, (2) complex spectral–spatial dependencies, (3) diverse anomaly scales, and (4) the need for computationally efficient inference. To address these challenges, we propose SM-HAD, a unified encoder–decoder framework that integrates three complementary modules. First, the Ortho Spectrum Fourier Block (OSFB) mitigates spectral redundancy by projecting features into the frequency domain and selectively enhancing informative spectral modes. Second, the Masked Vanilla Attention Block (MVAB) explicitly preserves local spatial relations through a context-aware masking mechanism, improving fine-grained anomaly localization. Third, the Residual Mamba Block (RMB) leverages state-space modeling to capture long-range spatial dependencies with linear complexity, enabling scalable inference. Extensive experiments and ablation studies on six benchmark datasets demonstrate that SM-HAD consistently outperforms 18 State-of-the-Art (SoTA) methods, achieving competitive accuracy while requiring fewer parameters and lower FLOPs. These properties make SM-HAD highly suitable for real-time and resource-constrained applications. The source code is available at https://github.com/Tanishq251/SM-HAD.

We introduce HyperCap, the first large-scale hyperspectral captioning dataset designed to enhance model performance and effectiveness in remote sensing applications. Unlike traditional hyperspectral imaging (HSI) benchmarks, HyperCap integrates spectral data with pixelwise textual annotations, enabling deeper semantic understanding. This dataset enhances model performance in tasks like classification and feature extraction, providing a valuable resource for advanced remote sensing applications. HyperCap is constructed from four benchmark datasets and annotated through a hybrid approach combining automated and manual methods to ensure accuracy and consistency. Empirical evaluations using state-of-the-art encoders and diverse fusion techniques demonstrate significant improvements in classification performance. These results underscore the potential of vision–language learning in HSI and position HyperCap as a foundational dataset for future research in the field. The code and dataset are available at https://github.com/arya-domain/HyperCap.

Traditional supervised 3D medical image segmentation models need voxel-level annotations, which require huge human effort, time, and cost. Semi-supervised learning (SSL) addresses this limitation of supervised learning by facilitating learning with a limited annotated and larger amount of unannotated training samples. However, state-of-the-art SSL models still struggle to fully exploit the potential of learning from unannotated samples. To facilitate effective learning from unannotated data, we introduce a novel approach LLM-SegNet, which exploits a large language model (LLM) to fuse task-specific knowledge into our co-training framework. This knowledge aids the model in comprehensively understanding the features of the region of interest (ROI), ultimately leading to more efficient segmentation. Additionally, to further reduce erroneous segmentation, we propose a Unified Segmentation loss function. This loss function reduces erroneous segmentation by not only prioritizing regions where the model is confident in predicting between foreground or background pixels but also effectively addressing areas where the model lacks high confidence in predictions. Experiments on Left Atrium and Pancreas-CT datasets demonstrate the superior performance of LLM-SegNet compared to the state-of-the-art.