Remote sensing imagery often involves complex backgrounds and multi-scale targets, while variations such as rotation and scaling significantly degrade the performance of existing object detection algorithms and hinder effective modeling of spatial relationships between objects. To address these challenges, we propose SFE-CapsNet. First, we fuse scalar features extracted by convolutional neural networks (CNNs) with vector features from capsule networks through structural reorganization to form virtual capsules, approximating the dynamic routing process via a single fully connected layer. This design preserves object pose and texture information while streamlining information flow. Second, we introduce a capsule attention module that generates attention masks to dynamically enhance target-relevant features and suppress background noise, strengthening multi-level feature representations. Integrated with a feature pyramid network (FPN) architecture, our approach achieves precise detection of targets at varying scales. Experimental results demonstrate that multi-level feature fusion and the capsule attention mechanism significantly improve detection accuracy and robustness, achieving 77.65% mean Average Precision (mAP) on the DOTA dataset and 97.63% mAP on HRSC2016, highlighting its effectiveness and efficiency in complex scenes.

Efficient fusion of infrared and visible images is of critical importance for real-time applications such as autonomous driving. While deep learning-based fusion methods have demonstrated significant improvements in fusion quality in recent years, current network architectures still exhibit unsatisfactory computational complexity and processing speed. To reduce computational complexity and improve fusion efficiency, mask-based methods or approaches driven by downstream tasks often prioritize key regions. However, such methods tend to overemphasize target objects, potentially overlooking contextually significant elements. To address this limitation and achieve more effective fusion, we propose BEFuse, a decoupled two-stage training strategy with an end-to-end inference framework. BEFuse extracts shallow features and gradient information from images in the first stage via a cross-modal image segmentation subnetwork. In the fusion stage, we use the Hadamard product to map features to an implicit quadratic feature space, combining feature similarity and gradient mask information, allowing automatic adjustment of loss weights and improving fusion accuracy. Experiments on four datasets (MSRS, TNO, RoadScene, and M3FD) show that BEFuse outperforms existing methods in both fusion quality and computational speed.

Cross-view geo-localization (CVGL) aims to match images of the same location captured from different viewpoints, such as those captured by Unmanned Aerial Vehicles (UAVs) and satellite platforms. The task is particularly challenging due to significant variations in scale, viewpoint, and illumination. Most existing methods employ symmetric sampling strategy to construct drone–satellite image pairs for deep metric learning, but neglect the potential of incorporating multi-view drone images to enhance the viewpoint robustness of features. To address this, we propose leveraging multi-view images to learn Domain-Invariant Discriminative Embeddings (DIDE) for CVGL. DIDE introduces an Inter-view Feature Aggregation Module (IFAM), which dynamically integrates multi-view drone information into robust embeddings. These are used in contrastive learning with satellite embeddings within batches to learn view-invariant discriminative features, while representation learning further improves scene discrimination across batches. To reduce the domain gap, DIDE constructs and aligns drone and satellite prototypes for effective cross-domain feature alignment. Furthermore, we adopt a parameter-efficient transfer learning strategy that leverages the capabilities of pre-trained foundation models while fine-tuning only dual adapters, significantly reducing the trainable parameters. DIDE achieves the state-of-the-art on University-1652 and University-160k, competitive results on SUES-200, and demonstrates strong cross-dataset transferability, with fewer training parameters and lower computational cost.

In partial label learning (PLL) tasks, each training instance is assigned a candidate label set, with only one label being correct. Previous studies on PLL have focused on scenarios where the class label set remains fixed, i.e., the label set for test data is the same as that used during training. However, in many real-world applications, the environment is dynamic, and new labels may emerge, requiring methods that can detect and classify these new labels. Moreover, previous methods typically learn from partial label data by manipulating the same feature set, which may be suboptimal as it overlooks the semantic relationships between instances and labels. To this end, we develop a novel PLL approach called Label-Specific feature-based Partial label learning with Emerging new Labels (LSPEL), which works by iteratively learning label-specific features during the label disambiguation process to support new label detection and model update. It consists of three key components: (1) model training based on label-specific feature learning, (2) construction of a new label detector that works in conjunction with the classifier to predict known labels, and (3) model updating and induction to further enhance the prediction results for known labels. Extensive experiments on synthetic and real-world PL datasets demonstrate that LSPEL is effective in handling emerging new labels.

Recently, contrastive learning approaches have achieved significant empirical success in representation learning for multivariate time series classification (MTSC). A key component of contrastive learning is to select appropriate augmentations. However, it remains an open question to find the desired augmentations of time series data for given datasets. Meanwhile, the existing augmentation methods focus on the design of feasible positive pairs while neglecting the construction of discriminative negative pairs. To address these issues, we propose a mixed supervised contrastive learning framework, called NI-MTSC, based on the neighborhood interpolation for MTSC. It integrates a novel neighborhood interpolation time series data augmentation and mixed supervised contrastive loss (MixCon). Firstly, the proposed neighborhood interpolation data augmentation method is used to generate more standardized and feasible positive and negative pairs for contrastive learning and form more distinct class boundaries in the representation space. The main steps include: i) we first use the distance metrics to explore the similarity of each sample data in different degrees and find the nearest-neighbor samples; ii) we adopt the linear interpolation to create new augmented samples between the original sample and the neighbor samples; iii) we introduce the label information of the original sample and the neighbor samples to construct positive and negative pairs. Secondly, to maximize the use of labels, we introduce the designed loss function MixCon, which combines the inter-class supervised and intra-class self-supervised contrastive loss to capture time series information hierarchically at the timestamp level for high-quality representations learning. The experiments on the 18 public datasets from the UEA MTSC archive and show that our proposed NI-MTSC achieves the best performance in classification accuracy compared with the state-of-the-art methods, indicating the effectiveness of our proposed method for the MTSC task.

Accurate multi-energy load forecasting is crucial for the optimal operation of Integrated Energy Systems (IES). This study innovatively proposes a hybrid prediction model with two core innovations: (1) We design a periodic Huber loss function that dynamically adjusts penalty weights, effectively balancing load periodicity characteristics with outlier robustness; (2) We propose a stacked prediction architecture that combines the dilated convolution properties of Temporal Convolutional Network (TCN) with the bidirectional temporal modeling capabilities of Bidirectional Gated Recurrent Unit (BiGRU). By connecting multiple fundamental modules in series through residual connections, the proposed model achieves progressive extraction and prediction of multi-scale temporal features in multi-energy load sequences through gradual residual output optimization. Experiments conducted on the actual operational data from Arizona State University’s campus energy system demonstrate that the proposed model exhibits better performance than the advanced methods (Informer and FEDformer). The experimental results indicate the proposed model is effective for real-time operational scheduling of integrated energy systems.

In controlled settings, facial age estimation models have achieved remarkable accuracy. However, in the uncontrolled or "in the wild" scenarios, the performance of these models faces significant challenges. Factors such as varying image resolutions, environmental lighting conditions, and obstructions substantially impact the effectiveness of age estimation in these less constrained environments. This complexity highlights the ongoing need for advancements in the field to address the unique challenges presented by the real-world conditions. To solve this problem, we propose a novel model that combines Convolutional Neural Networks (CNN) and Vision Transformer architectures. We have made a strategic modification to the ResNet architecture, which forms the backbone of our model. A significant enhancement in our model is the replacement of traditional convolutions with depth-wise separable convolutions, which offers improved efficiency and reduced model complexity. In addition, we replace the last three bottleneck layers with the components of the Vision Transformer. This alteration allows for a more sophisticated and nuanced feature extraction process, particularly beneficial for complex image scenarios encountered in the uncontrolled settings. Additionally, our model incorporates a feature pooling pyramid structure to facilitate the multi-scale feature fusion. This innovative combination not only leverages the strengths of both CNN and Transformers, but also enhances the feature extraction and representation capabilities across various scales, making it particularly effective for diverse vision-based tasks. We evaluate our model on three uncontrolled popular datasets and validate the effectiveness of the proposed method.

Student performance prediction is vital for identifying at-risk students and providing support to help them succeed academically. In this paper, we propose a feature importance-based multi-layer CatBoost approach to predict the students' grade in the period exam. The idea is to construct a multi-layer structure with increasingly important features layer by layer. Specifically, the feature importance are first calculated and sorted in ascending order. In each layer, features with the least importance are accumulated until reaching a given threshold. Then, these selected features are used to construct the first layer by training the CatBoost. Next, this trained CatBoost is utilized to generate a feature that adds to the feature set with their importance within a threshold. After that, all these feature are used to train the CatBoost in the next layer. This process is repeated until all the features are used. The results show that the proposed model has the best performance. Moreover, the statistical test conducted based on 20-runs of experiments validates the significant superiority of our proposed model over the compared models and demonstrates the efficacy of the multi-layer structure in enhancing the proposed model. This indicates our proposed model can help decision makers in enhancing educational quality.

To address the limitations of existing house price prediction methods that struggle with handling mixed data types, we developed a novel model utilizing a reinforcement learning-based weight fusion algorithm. First, we reconstruct the feature fusion vectors from categorical feature variables. Then, we create individual house price prediction models using a variety of algorithms: Multiple Linear Regression, Regression Decision Tree, Extreme Random Regression Tree, and Multilayer Perceptron. These models' weights are dynamically adjusted using Q-Learning to minimize the Root Mean Square Error (RMSE). Our experiments demonstrate substantial gains in RMSE performance using our weight fusion approach. Specifically, on the small-scale Shenzhen dataset, RMSE improvements vary from 8.65% to 64.61%. On the large-scale Boston dataset, the improvements range between 53.06% and 76.40% in comparison to the standalone prediction algorithms.

Although machine learning classifiers have been successfully used in the medical and engineering fields, there is still room for improving the predictive accuracy of model classification. The higher the accuracy of the classifier, the better suggestions can be provided for the decision makers. Therefore, in this study, we propose an ensemble machine learning approach, called Feature generation-based Ensemble Support Vector Machine (FESVM), for classification tasks. We first apply the feature selection technique to select the most related features. Next, we introduce an ensemble strategy to aggregate multiple base estimators for the final prediction using the meta-classifier SVM. During this stage, we use the classification probabilities obtained from the base classifier to generate new features. After that, the generated features are added to the original data set to form a new data set. Finally, this new data set is utilised to train the meta-classifier SVM to obtain the final classification results. For example, for a binary classification task, each base classifier has two probabilities (p for one class and 1−p for the other class). In this case, two new features are generated from the combination of probabilities based on these base classifiers. One is the sum of p as new feature 1, and the other is the sum of 1−p as new feature 2. These two new features are then added to the original data set to form the new data set. In the same way, our feature generation method can be easily extended for a multi-class task for generating new features, where the number of features depends on the number of classes. Those generated features from the base estimators (first layer) are added to the original data set to form a new data set. This new data set is used as the input to the second layer (meta-classifier) to obtain the final model. Experiments based on the 20 data sets show that our proposed model FESVM has the best performance compared to the other machine learning classifiers under comparison. In addition, our FESVM has better performance than the original stacking method in the multi-class classification tasks. Statistical results based on the Wilcoxon–Holm method also confirms that our FESVM can significantly outperform the other models. These indicate that our FESVM can be a useful tool for classification tasks, especially multi-classification tasks.

DBSCAN is a classical clustering algorithm, which can identify different shapes and isolate noisy patterns from a dataset. Despite the above advantages, the bottleneck of DBSCAN is its computation time for high dimensional datasets. This work, thus, presents a simple and fast method to improve the efficiency of DBSCAN algorithm. We reduce the execution time in two aspects. The first one is to reduce the number of points presented to DBSCAN and the second one is to apply the HNSW technique instead of the linear search structure for improving its efficiency. The experimental results show that our proposed algorithm can greatly improve the clustering speed without losing or even obtaining better accuracy, especially for large-scale datasets.

Collaborative filtering is a popular tool for recommendation systems. However, collaborative filtering technologies often suffer from high time complexity, the cold-start problem, and low coverage. Recent research shows that social networks and trust-aware methods can effectively solve these proble ms. Therefore, we propose a Trust Domain Expert Collaborative Filtering recommendation system. First, we divide the user item rating matrix into multiple sub-matrices based on the domain attributes of each item. For each sub-matrix, we then use domain experts to construct a user–expert trust matrix. Finally, combined with the target user’s domain of interest, we predict their missing ratings. Experimental results show that this method not only improves the accuracy and recommended coverage of collaborative filtering-based methods, but also reduces the computation time.