Graph neural networks (GNNs) are widely used in information retrieval, but they often require large amounts of labeled data. To address this problem, self-supervised methods like graph contrastive learning (GCL) are developed to learn from graph structures without labeled data. However, GCL faces a challenge in practice. Plenty of traditional GCL methods differ fundamentally from GNNs in handling neighboring nodes, hindering effective contrastive learning. To address the above issue, we propose a graph self-supervised learning model based on Contrastive Graph Matching (ConGM). The model effectively mitigates the conflict between GCL methods and the homophily assumption of GNNs by using linear node matching and quadratic edge alignment mechanisms to treat some neighboring nodes as both positive and negative samples, rather than considering all neighboring nodes as negative samples as in traditional GCL methods. Additionally, to tackle the imbalance of positive and negative samples in edge alignment, we design a bi-level negative sample selection strategy to choose appropriate hard negative samples. Extensive experiments conducted on multiple benchmark datasets have validated the effectiveness of our proposed method.

Recommender Systems (RS) play a critical role in enhancing user experiences across online platforms by modeling user-item interactions as bipartite graphs. Predicting signed links in such graphs remains challenging due to the sparsity and complexity of sign distributions and the limitations of traditional methods like matrix factorization and Graph Convolutional Networks (GCNs), which often fail to capture the intricate local topological and sign-based patterns essential for accurate predictions. To address these challenges, we propose SMA-GNN, a framework specifically designed for signed link prediction in bipartite graphs. SMA-GNN combines Local Subgraph Extraction, Two-Anchor Distance Labeling (TADL), and a Symbol-aware Multi-head Attention Mechanism to enhance predictive capability and interpretability. By extracting a closed local subgraph around the target link, our method captures relevant topological and sign contexts. TADL refines this by assigning unique structural labels to nodes based on their proximity to anchor nodes, encapsulating roles and relationships. The symbol-aware attention mechanism integrates edge sign information into the message-passing process, generating highly discriminative subgraph embeddings. Experiments on benchmark datasets show that SMA-GNN outperforms global embedding methods in prediction accuracy and provides deeper insights into user-item interactions, enabling more precise and personalized recommendations. Our code is avilable at https://github.com/xiaohuzidefeijian/SMAGNN/tree/master

Link sign prediction in signed bipartite graphs, which are extensively utilized across diverse domains such as social networks and recommendation systems, has recently emerged as a pivotal challenge. However, significant space and time complexities associated with the scalability of bipartite graphs pose substantial challenges, particularly in large-scale environments. To address these issues, this paper introduces the SignFlow Bipartite Subgraph Network (SBSN), balancing sublinear training memory growth through a heuristic subgraph extraction method integrated with a novel message passing module, with optimal inference efficiency achieved via the node feature distillation module. Our subgraph sampling approach reduces the graph size by focusing on neighborhoods around target links and employs an optimized directed message passing mechanism to aggregate critical structural patterns. This mechanism allows SBSN to efficiently learn rich local structural patterns essential for accurate sign prediction. Furthermore, to overcome the inefficiency of subgraph sampling-based models during inference, SBSN incorporates a node feature distillation module after the first training stage. This module distills subgraph features into node features, enabling fast inference while retaining the rich structural information of subgraphs. Experiments reveal that SBSN shows superior performance in both medium- and large-scale datasets, efficiently managing memory and computational resources, making it a scalable solution for extensive applications.

Signed bipartite graphs are widely employed in social networks, recommender systems, and other domains, where the nodes represent two different sets, such as users and commodities, and the links have positive and negative signs to reflect the ratings. Link sign prediction is a crucial task, but the link class imbalance when one type of link (e.g., the head class) is significantly more numerous than another (e.g., the tail class) makes this task extremely challenging. To address this challenge, we propose a Line-Graph-Based Dynamic Balancing Prediction (LDBP) method. Specifically, we first convert the links of a signed bipartite graph into nodes of a line graph, and then the link class imbalance problem in the bipartite graph is naturally transformed to the problem of node class imbalance in the line graph. To tackle this problem, we introduce a Centroid Contrastive Learning (CCL) method and design a Dynamic Synthesis & Deletion (DSD) strategy for the tail-class nodes. By dynamically adjusting the quantity of synthesized tail-class nodes, we obtain a Balanced Line Graph (BaLG). Extensive experiments on real-world datasets demonstrate the effectiveness of our method in improving the accuracy of link sign prediction and addressing the issue of link class imbalance.

Personalized news recommendation plays a vital role in mitigating information overload, yet challenges persist in accurately capturing user preferences and fine-grained interests. Leveraging the semantic understanding and extraction capabilities of large language models (LLMs), we propose a Multi-Interest Personalized News Recommendation (MIPNR) model to address these issues. MIPNR separately models user interests at the user, news, and entity levels. Specifically, we introduce a Category-Guided Interest-News Matching (CGIN-Matching) method to identify potential interests, a Local News Entity Graph (LNEG) to model subtle entity relationships, and an entity-wise attention mechanism to extract fine-grained interests. In addition, LLMs are used to generate explicit textual descriptions of user preferences. Extensive experiments on real-world datasets demonstrate the effectiveness of our approach.

In the field of graph representation learning for molecular property prediction, self-supervised learning is highly appreciated for its ability to mine potential connections in unlabeled data. However, existing approaches such as neighborhood prediction or contrastive learning often lack in fusing structural and semantic features of molecular graphs. To address this issue, we propose a Graph Matching Based Graph Self-Supervised Learning (GMSSL) framework, which aims to comprehensively learn the structure and semantics of molecular graphs. Specifically, utilizing intra-graph and inter-graph convolution and graph matching methods, GMSSL is able to encode complex information for efficient training without supervision. After pre-training and fine-tuning, GMSSL demonstrates powerful molecular property prediction on multiple datasets.

Large Vision Model (LVM) has recently demonstrated great potential for medical imaging tasks, potentially enabling image enhancement for sparse-view Cone-Beam Computed Tomography (CBCT), despite requiring a substantial amount of data for training. Meanwhile, Deep Image Prior (DIP) effectively guides an untrained neural network to generate high-quality CBCT images without any training data. However, the original DIP method relies on a well-defined forward model and a large-capacity backbone network, which is notoriously difficult to converge. In this paper, we propose a stable optimization method for the forward-model-free, LVM-based DIP model for sparse-view CBCT. Our approach consists of two main characteristics: (1) multi-scale perceptual loss (MSPL) which measures the similarity of perceptual features between the reference and output images at multiple resolutions without the need for any forward model, and (2) a reweighting mechanism that stabilizes the iteration trajectory of MSPL. One shot optimization is used to simultaneously and stably reweight MSPL and optimize LVM. We evaluate our approach on two publicly available datasets: SPARE and Walnut. The results show significant improvements in both image quality metrics and visualization that demonstrates reduced streak artifacts. The source code is available upon request.

To enhance the accuracy and reliability of the response of large language models (LLMs) in professional domain-specific question answering, this paper proposes an approach to enhance the capabilities of LLMs by prompting them to "think twice." This methodology encourages a second-pass processing, allowing for deeper analysis and more refined outputs. We developed a prototype-logic-inspired query optimizer (PQO) to enhance the response of LLM. It refines initial queries using few-shot learning and self-reflection, leading to more insightful questions. Our experimental results show that the proposed approach obtains empirically validated answers.

Meta-learning methods have shown to be effective in dealing with cold-start recommendation. However, most previous methods rely on an ideal assumption that there exists a similar data distribution between source and target tasks, which are unsuitable for the scenario that only extremely limited number of new user or item interactions are available. In this paper, we propose to boost meta-learning cold-start recommendation with graph neural network (MeGNN). First, it utilizes the global neighborhood translation learning to obtain consistent potential interactions for all new user and item nodes, which can refine their representations. Second, it employs the local neighborhood translation learning to predict specific potential interactions for each node, thus guaranteeing the personalized requirement. In experiments, we combine MeGNN with two representative meta-learning models MeLU and TaNP. Extensive results on two widely-used datasets show the superiority of MeGNN in four different scenarios.

Deriving from image matching and understanding, semantic keypoint matching aims at establishing correspondence between keypoint sets in images. As graphs are powerful tools to represent points and their complex relationships, graph matching provides an effective way to find desired semantic keypoint correspondences. Recent deep graph matching methods have shown excellent performance, but there is still a lack of exploration and utilization of spatial information of keypoints as nodes in graphs. More specifically, existing methods are insufficient to capture the relative spatial relations through current graph construction approaches from the locations of semantic keypoints. To address these issues, we introduce a positional reconstruction encoder-decoder (PR-EnDec) to model intrinsic graph spatial structure, and present an end-to-end graph matching network PREGM based on PR-EnDec. Our PR-EnDec consists of a positional encoder that learns effective node spatial embedding with the affine transformation invariance, and a spatial relation decoder that further utilizes the high-order spatial information by reconstructing the locational structure of graphs contained in the node coordinates. Extensive experimental results on three public keypoint matching datasets demonstrate the effectiveness of our proposed PREGM.

Continual graph learning (CGL) aims to mitigate the topological-feature-induced catastrophic forgetting problem (TCF) in graph neural networks, which plays an essential role in the field of information retrieval. The TCF is mainly caused by the forgetting of node features of old tasks and the forgetting of topological features shared by old and new tasks. Existing CGL methods do not pay enough attention to the forgetting of topological features shared between different tasks. In this paper, we propose a transformer-based CGL method (Trans-CGL), thereby taking full advantage of the transformer's properties to mitigate the TCF problem. Specifically, to alleviate forgetting of node features, we introduce a gated attention mechanism for Trans-CGL based on parameter isolation that allows the model to be independent of each other when learning old and new tasks. Furthermore, to address the forgetting of shared parameters that store topological information between different tasks, we propose an asymmetric mask attention regularization module to constrain the shared attention parameters ensuring that the shared topological information is preserved. Comparative experiments show that the method achieves competitive performance on four real-world datasets.

CT vertebral segmentation plays an essential role in various clinical applications, such as computer‐assisted surgical interventions, assessment of spinal abnormalities, and vertebral compression fractures. Automatic CT vertebral segmentation is challenging due to the overlapping shadows of thoracoabdominal structures such as the lungs, bony structures such as the ribs, and other issues such as ambiguous object borders, complicated spine architecture, patient variability, and fluctuations in image contrast. Deep learning is an emerging technique for disease diagnosis in the medical field. This study proposes a patch‐based deep learning approach to extract the discriminative features from unlabeled data using a stacked sparse autoencoder (SSAE). 2D slices from a CT volume are divided into overlapping patches fed into the model for training. A random under sampling (RUS)‐module is applied to balance the training data by selecting a subset of the majority class. SSAE uses pixel intensities alone to learn high‐level features to recognize distinctive features from image patches. Each image is subjected to a sliding window operation to express image patches using autoencoder high‐level features, which are then fed into a sigmoid layer to classify whether each patch is a vertebra or not. We validate our approach on three diverse publicly available datasets: VerSe, CSI‐Seg, and the Lumbar CT dataset. Our proposed method outperformed other models after configuration optimization by achieving 89.9% in precision, 90.2% in recall, 98.9% in accuracy, 90.4% in F‐score, 82.6% in intersection over union (IoU), and 90.2% in Dice coefficient (DC). The results of this study demonstrate that our model’s performance consistency using a variety of validation strategies is flexible, fast, and generalizable, making it suited for clinical application.

Recently, self-supervised representation learning gives further development in multimedia technology. Most existing self-supervised learning methods are applicable to packaged data. However, when it comes to streamed data, they are suffering from catastrophic forgetting problem, which is not studied extensively. In this paper, we make the first attempt to tackle the catastrophic forgetting problem in the mainstream self-supervised methods, i.e., contrastive learning methods. Specifically, we first develop a rehearsal-based framework combined with a novel sampling strategy and a self-supervised knowledge distillation to transfer information over time efficiently. Then, we propose an extra sample queue to help the network separate the feature representations of old and new data in the embedding space. Experimental results show that compared with the naive self-supervised baseline, which learns tasks one by one without taking any technique, we improve the image classification accuracy by 1.60% on CIFAR-100, 2.86% on ImageNet-Sub, and 1.29% on ImageNet-Full under 10 incremental steps setting. Our code will be available at https://github.com/VDIGPKU/ContinualContrastiveLearning.