Amajor assumption in data mining and machine learning is that the training set and test set come from the same domain.Theysharethe samefeature spaceandhavethe same distribution.However,in manyreal-world applications,the training set and test set usually come from different domains. Thus, there might be negative similarities between different domainsso thatthe negative transferproblem causedbynegative similaritymayhappen.In this paper, weproposeanovel method named active transfer learning (ATL) to solve the above problem. Speci?cally, the orthogonal projection matrix and theweight coef?cient vectorareintroducedtoextendmaximum mean discrepancy(MMD)sothatitcan minimizeMMDand simultaneouslyeliminate the negative transfer.To ?nd the informative and discriminative subsets from the source domain, we thenpropose an information diversity termbyusing the localgeometric structure informationof source samples. Besides,byusing the label information of source samples, our method can guarantee the selected subsets as discriminative as possible. Finally, to ef?cientlyimplement the proposed method, an alternating optimization approach, whichis based on the alternating direction methodofmultipliers (ADMM),is designedto solvethe optimizationproblem.To demonstratethe effectivenessoftheproposedATL model,theexperiments are conductedon ?ve real-world data sets.Theexperimental results show the superiority of our method over the state-of-the-art methods.

Community detection in graph networks is a fundamental yet challenging task due to limitations in existing methods. Traditional Graph Convolutional Network (GCN)–based models often suffer from over-smoothing, which results in indistinguishable node representations after excessive aggregation. Additionally, many models face computational inefficiency, restricting their scalability on large-scale networks. To address these challenges, we propose GATFELPA, a hybrid community detection model that combines a Graph Attention Network (GAT) with an enhanced label propagation algorithm (DPCELPA). GATFELPA employs an adaptive strategy to dynamically determine the optimal number of aggregation layers, effectively mitigating over-smoothing by balancing intra-community compactness and inter-community separability. A novel similarity preservation module further enhances the model’s ability to differentiate communities by retaining local and global dissimilarities in heterogeneous networks. Comparative experiments on four real-world datasets, including large-scale networks like ogbn-arxiv, demonstrate that GATFELPA achieves superior performance across most metrics, particularly excelling in accuracy, robustness, and scalability. These results highlight GATFELPA as a promising approach for addressing complex community detection tasks.

To accurately retrieve similar objects from different domains, domain adaptive retrieval method is applied to cope with the domain shift problem in information retrieval. However, existing methods still have two problems: a) they fail to filter out low-confidence samples, leading to error accumulation; and b) they ignore the negative effect of domain discrepancy. To address these two issues, we propose an efficient method called Dynamic Confidence Sampling and Label Semantic Guidance Learning (DCS-LSG). First, Dynamic Confidence Sampling (DCS) is employed to dynamically select high-confidence samples from the target domain so as to improve the effectiveness of learning. Second, Label Semantic Guidance (LSG) learning is presented to enhance the label semantics of features during domain adaptive retrieval. In addition, we introduce a Dual-Projection Relaxation (DPR) strategy to learn more effective features on two specific projection spaces. At last, a two-step hashing strategy is used to generate high-quality hash codes. Experiments on multiple cross-domain retrieval datasets demonstrate that the proposed DCS-LSG can achieve a significant performance improvement. The source code is available at https://github.com/Kybin-kb/DCS-LSG-code/tree/master .

Due to high computational efficiency and low storage cost, cross-modal hashing retrieval attracts much attention. However, as heterogeneous data from different modalities often have distinct physical meanings and underlying structures, learning encoding with equal length for different modalities may result in an insurmountable semantic gap. In addition, there are still some issues, e.g., how to combine label and sample information to learn hash codes effectively, how to reduce the time consumption caused by computing n × n similarity matrix, and how to effectively solve the complex discrete optimization problem. To overcome the above challenges, this study propose a novel model called Scalable Discrete and Asymmetric Unequal Length Hashing (SDAULH). First, SDAULH constructs a novel hash model that utilizes unequal length encoding schemes to narrow the semantic gap between heterogeneous modalities. Second, SDAULH develops a dual semantic embedding learning scheme, which combines pairwise similarity between label and sample data to generate a more discriminative hash code. Third, SDAULH associates with both hash codes and label information by an asymmetric relaxation strategy. Furthermore, SDAULH solves directly the discrete optimization problem by generating discrete hash codes. Experimental results on four benchmark datasets demonstrate the promising performance of SDAULH.