To overcome limitations of traditional species distribution models-particularly their over-reliance on environmental variables and inadequate integration of biological-geospatial data-we propose CGAF-SDP (Cross-Graph Attention Fusion framework for Endangered Species Distribution Prediction), the first GNN-based framework for this task. Our dual-graph approach combines: (1) a homogeneous species-region associ-ation graph, and (2) a heterogeneous graph that integrates species features (extracted through named entity recognition) with geospatial embeddings. Key innovations comprise: (1) A cross-graph attention mechanism enabling multi-source fusion; (2) Inductive matrix completion addressing unobserved taxa prediction; (3) The Species-Area Relationship Database (SADB) for systematic evaluation. Experimental results validate CGAF-SDP's superior performance, highlighting GNNs' potential for ecological modeling.

For inductive link prediction in knowledge graphs, we address the problem by considering bridging links (i.e., links connecting discrete graphs). Although current research overcomes the traditional graph topological constraints, it tends to ignore the dynamic interactions between relations and entities as well as the capture of global information. The problem of discrete and small amounts of information in real-world knowledge graphs makes it difficult for existing methods to effectively integrate global and local information and to model complex relationships between entities. To address these issues, we propose a novel meta-analogy learning framework, Ank-motor, which integrates autonomously designed dynamic graph neural networks with analogical reasoning. The dynamic graph neural network models interactions and captures global information, while the analogical inference layer integrates entity, relation, and triple-layer information to capture local semantic details. In addition, meta-learning techniques are utilized to deal with problems with small amounts of data and to enhance the model’s ability to generalize to new tasks. Numerous experiments show that Ank-motor significantly outperforms existing models on multiple benchmark datasets.

Traditional knowledge graph completion(KGC) approaches require a large number of training instances and hold a closed-world assumption. The real case is that very few instances are available and evolve quickly with the new entities and relations being added by the minute. So we propose a zero or few shot knowledge graph completions task and design a new joint embedding model JoinE by text enhancement with multi-grained attention. More specifically, a probabilistic version of TransE with logistic loss is firstly used to learn triples-based embedding from the KG. Then, multi-grained attention mechanism is used to capture text relevance. The intra-attention is multi-head self-attention, which enhances relevant word within single text knowledge. The inter-attention minimizes noise or information loss in multiple text knowledge integration based on the triple-dependent principle. Finally, a gate mechanism is used to automatically fuse triples-based embedding and text-based embedding. When the gate is close to 0, the corresponding embedding vector is ignored. Otherwise, it will become more important. Experimental results on few shot and zero shot task demonstrate that the proposed approach can outperform the state-of-the-art algorithms.