Diffusion-based image captioning models effectively mitigate the token dependency issue inherent in autoregressive methods. However, the noise introduced in diffusion methods weakens sentence information, resulting in insufficient ability of image-text feature alignment. To address this issue, we propose a multipath alignment learning for image captioning with feature mask (MALFM-Captioner) method. Leveraging both global and regional visual features, we first introduce a feature masked module (FMM) that enables the model to reconstruct masked visual information during training, thereby enhancing its capability to learn discriminative visual representations. Concurrently, we perform cross-attention between deep image and text features and fuse the outputs via weighted summation, effectively mitigating the image-text misalignment problem inherent to single-path paradigms. Furthermore, we design a concise yet effective gated feature fusion module (GFFM) to integrate complementary feature alignment results, improving the accuracy and semantic fidelity of generated captions. Extensive experiments on the MS COCO and Flickr 30K datasets demonstrate that MALFM-Captioner achieves 0.9% and 1.9% improvements in Bleu-4 and CIDEr metrics, respectively, and exhibits competitive performance against state-of-the-art models (e.g., DDCap and Bit Diffusion).

Designing effective graph neural networks (GNNs) for diverse tasks requires substantial manual effort, especially when dealing with the intricate interplay between topological structures and semantic information in graph data. Existing automated approaches often rely on off-the-shelf model combinations that remain superficial, resulting in generated model frameworks that are hardly interpretable and limited by the capabilities of the combined models. To address this gap, we propose an evolutionary-based framework that systematically searches for optimal GNN architectures while adaptively combining topological and semantic insights. This approach reduces human intervention, accommodates a wide range of graph tasks, and ensures robust performance in complex scenarios. Extensive experiments on real-world datasets demonstrate that our method outperforms both handcrafted GNNs and mainstream AutoML models in terms of accuracy, offering a scalable and adaptive solution for GNN design.

Network motifs, as fundamental functional substructures in gene regulatory networks (GRNs), play a critical role in regulating gene expression. Despite the successful application of graph representation learning in GRN modeling, most existing approaches mainly capture pairwise relationships and overlook higher order regulatory patterns encoded by functional motifs, which limits the accuracy of regulatory inference. To address this limitation, we propose Motif-GRN, a motif-based hypergraph representation learning framework that captures the underlying biological logic in higher order semantic structures. We first identify statistically significant regulatory motifs and construct a multichannel motif-induced hypergraph. We then design a motif-aware hypergraph convolutional network to extract motif-centric semantic features, while a conventional graph convolution module preserves first-order relational information. In addition, we introduce cross-view contrastive learning to align heterogeneous representations and enhance gene embeddings. Building on Motif-GRN, we develop an inductive extension that enables cross-dataset generalization and effective GRN inference with limited labels. Extensive experiments on three ground-truth networks across seven cell types demonstrate that Motif-GRN outperforms state-of-the-art baselines in both transductive and inductive GRN inference tasks, highlighting its potential for higher order regulatory network modeling.

Recent advancements in weakly supervised semantic segmentation (WSSS) have shown promise by using the contrastive language-image pretraining (CLIP) model to generate pseudo-labels. However, directly applying the CLIP model without considering interclass relationships in downstream tasks has resulted in suboptimal transferability and generalization. To address these challenges, we propose CLIP graph adapter (CLIP-GA), a novel approach that integrates both textual and visual structural knowledge to generate high-quality initial class activation maps (CAMs) for each object class. Our method introduces a dual-graph adaptive strategy, comprising a textual subgraph and a visual subgraph and employs cross-modal graph attention (CGA) for effective fusion. The framework includes three specialized loss functions that help to capture more complete object regions while minimizing the activation of background areas closely related to foreground objects. In addition, we implement the superpixel consistency to refine pseudo-labels and introduce a graph reasoning attention (GRA) module to build global contextual relationships within visual features for the segmentation network. Extensive experiments on the PASCAL VOC 2012 and MS COCO 2014 datasets have convincingly demonstrated the effectiveness of CLIP-GA compared with other state-of-the-art methods. Our code is provided at: https://github.com/JIA-ZHANG666/CLIP-GA.

In node classification, traditional graph neural networks (GNNs) typically assume implicit homophily, indicating that intraclass nodes are likely connected. However, real-world graphs frequently exhibit heterophily, in which interclass nodes are also commonly connected. To address this challenge, recent methods have adopted approaches such as expanding local neighborhoods and employing adaptive message aggregation to enhance the GNN performance on heterophily graphs. Nevertheless, these methods are restricted by the homophily assumption and fail to effectively capture long-range dependencies (e.g., widely separated intraclass nodes) and insufficiently leverage the graph topology. This study investigates the performance differences of GNN when it is applied to both homophily and heterophily graphs and finds that the distinguishability of neighborhood label distributions (NLDs) exhibits a significant correlation with the accuracy of node classification. To assess the impact of NLD on node classification, this study proposes a novel homophily metric based on node distinguishability. Subsequently, this study introduces a new GNN model named NLD-based GNN (NLDGNN) for node classification. First, NLDGNN initializes node representations by integrating node features with node NLDs. To address long-range dependencies in heterophily graphs, NLDGNN utilizes the global label relationship matrix with low-rank characteristics for global message passing. By combining the attention scores derived from the initial node representations, NLDGNN constructs the global label relationship matrix for enhanced message passing, thereby improving the expressiveness of node representations. Experimental results indicate that NLDGNN outperforms existing GNN models on both real-world homophily and heterophily graphs. The code of this study is available at https://github.com/wanli6/NLDGNN.

Graph Convolutional Networks (GCNs) have exhibited considerable promise in 3-D skeleton-based human motion prediction. Based on the intuitive observation that human motion can be delineated through the physical interconnections among human joints, many previous works have designed multiscale graphs to learn the relationships and constraints between different graph scales, obtaining encouraging results for human motion prediction. However, these fixed multiscale graphs obtain new scale graphs by merging adjacent human joint information, ignoring implicit semantic information during dynamic movements. Furthermore, human joint correlations tend to vary randomly as the depth of the multiscale clustering graph increases, which contradicts the design concept of fixed multiscale graphs. To address these limitations, we explore a novel correlation-based multiscale graph clustering network (CMGC) for adaptive multiscale graph representation learning. Given a human joints graph, the goal of CMGC is first to generate more new graphs representing motion correlations adaptively at different scale levels and then selectively restore the derived graph scales to the original human joints graphs, which enables various motion features extraction. Moreover, we introduce the discrete wavelet transform (DWT) to compensate for the signal loss caused by discrete cosine transform (DCT) domain modeling from human motion. The CMGC gives rise to gratifying performances with the adaptive multiscale graph. Extensive experiments reveal that CMGC outperforms state-of-the-art methods by 11.2%, 10.1%, and 11.2% of 3-D mean per joint position error (MPJPE) on average on Human 3.6M, CMU Mocap, and 3DPW datasets, respectively. We also test the mean angle error (MAE) on Human3.6M, which is lower by 6.5% than previous methods. Our code is released at https://github.com/JunyuShi02/CMGC.

Graph neural networks (GNNs) have achieved strong results on homophilic graphs with complete node attributes, yet their performance significantly deteriorates when faced with the combined challenges of heterophily and feature missingness. Heterophily introduces semantic inconsistency in neighborhoods, while feature missingness obscures node identity, which together constitute a complex problem we define as the heterophily-missing coupling (HMC). Under HMC, information exchanged between nodes becomes less reliable, and the usual assumptions that support message propagation no longer hold. To address this, we propose a novel adaptive prototype-guided personalized propagation (APP) framework. Specifically, it first leverages semantic rectification via prototypes (SRPs) to align neighborhood information with prototype semantics, reducing noise from inconsistent neighbors. Subsequently, personalized virtual propagation (PVP) builds upon this by clustering to construct prototype-aligned virtual edges, enabling effective feature imputation through minimizing Dirichlet energy across both real and virtual graphs. Finally, adaptive representation synergy (ARS) consolidates the propagated and imputed features by employing prototype-guided confidence weighting and enhancing representation quality via a contrastive training objective. Extensive experiments on multiple benchmark datasets demonstrate that APP consistently improves node classification performance on heterophilic graphs with missing features, achieving up to 11.22% improvement over state-of-the-art baselines while significantly reducing imputation error. The implementation is publicly available at https://github.com/limengran98/APP.

Feature selection is crucial for fuzzy decision systems (FDSs), as it identifies informative features and eliminates rule redundancy, thereby enhancing predictive performance and interpretability. Most existing methods either fail to directly align evaluation criteria with learning performance or rely solely on nondirectional Euclidean distances to capture relationships among decision classes, which limits their ability to clarify decision boundaries. However, the spatial distribution of instances has a potential impact on the clarity of such boundaries. Motivated by this, we propose spatially-aware separability-driven feature selection (S2FS), a novel framework for FDSs guided by a spatially-aware separability criterion. This criterion jointly considers within-class compactness and between-class separation by integrating scalar-distances with spatial directional information, providing a more comprehensive characterization of class structures. S2FS employs a forward greedy strategy to iteratively select the most discriminative features. Extensive experiments on 11 real-world datasets demonstrate that S2FS consistently outperforms ten state-of-the-art feature selection algorithms in both classification accuracy and clustering performance, while feature visualizations further confirm the interpretability of the selected features.

The cooperative multiagent reinforcement learning (MARL) has been widely used in many practical applications. Despite its success, a fundamental issue arises in MARL that agents face the dilemma of whether to select the best action to maximize rewards or to acquire more information collectively by exploring the novel states/actions due to partial observability. To solve this issue, existing methods merge exploration and exploitation methods. However, these methods are always suboptimal and may lead to failure in finishing tasks. In this article, we theoretically prove the existence of a latent state that can guarantee the optimal individual and global policies. Moreover, we prove that such a latent state can be approximately obtained by local observations. Based on the analysis, we propose a method named unified MARL (UMARL), which is a weighted value function factorization approach unifying exploitation and exploration in one framework. Specifically, we design the agent representation network (ARN) and individual weighting networks (IWNs) to learn agents' unified representations and weights of credit. Moreover, a latent state regularizer (LSR) is designed to encourage agents' representations to approximate the latent state. Extensive experiments show that UMARL can achieve superior performance compared with 12 state-of-the-art methods on $m$ -step matrix game, level-based foraging (LBF), StarCraft II, and Google research football (GRF). The source code is available at: https://github.com/CrazyBayes/UMARL.