graph neural network
PulseAugur coverage of graph neural network — every cluster mentioning graph neural network across labs, papers, and developer communities, ranked by signal.
10 天有情绪数据
-
GNN分析显示欧盟碳机制转移电力价格
本文介绍了一个图神经网络（GNN）框架，用于分析欧盟碳边境调节机制（CBAM）对电力价格和碳强度的影响。该研究对八个欧洲国家的子图进行了建模，揭示了CBAM创造了结构性市场差异，而不是作为一种统一的税收。结果表明，法国和瑞士等低碳国家由于竞争优势可能导致国内价格下降，而波兰等高碳国家则面临成本增加。
-
AI simulates VC collective decision-making to predict startup success
Researchers have developed a new collective agent system called SimVC-CAS to predict startup success by simulating venture capital decision-making as a multi-agent interaction process. This system utilizes role-playing …
-
GEM framework boosts dialogue state tracking with graph-enhanced experts and ReAct agents
Researchers have developed GEM, a novel framework for Dialogue State Tracking that combines graph-enhanced mixture-of-experts with ReAct agents. This approach dynamically routes between specialized experts, including a …
-
图神经网络增强量子架构搜索和无线系统
研究人员开发了一种新颖的图神经网络（GNN）方法，用于优化具有可重构智能表面的多基站系统的频谱和能源效率。另外，一种受AlphaGo启发并利用GNN的新型魔术信息量子架构搜索技术，旨在控制量子资源以改进电路设计。
-
GNNs create hierarchy-aware knowledge graph embeddings for yeast phenotype prediction
Researchers have developed a novel method using graph neural networks (GNNs) to create hierarchy-aware embeddings for knowledge graphs. This approach incorporates semantic loss derived from ontologies to better represen…
-
大型语言模型通过文本归因知识图谱增强医学概念表示
研究人员开发了MedCo框架，该框架利用大型语言模型来增强知识图谱中的医学概念表示。该方法通过推断缺失的关系并整合文本中的丰富语义信息，解决了现有医学本体的局限性。MedCo生成节点描述和边解释，将文本语义与图结构融合，创建统一的概念嵌入，从而改进下游临床预测任务。
-
联邦GNN通过隐私保护半监督学习提升GDM预测能力
研究人员开发了一种新颖的联邦半监督学习框架FedTGNN-SS，用于在保护医院间数据隐私的同时预测妊娠期糖尿病（GDM）。该方法通过使用原型引导伪标签和自适应图细化来解决标记数据有限和无法共享患者记录的挑战。在三个数据集上的实验表明，FedTGNN-SS的有效性，尤其是在标记稀缺性高的情况下，相比现有联邦方法取得了显著改进。
-
PIEGraph combines physics and GNNs for data-efficient robotic object dynamics
Researchers have developed PIEGraph, a new method that combines analytical physics with equivariant graph neural networks to learn object dynamics from limited interaction data. This approach improves the physical feasi…
-
SCGNN introduces granular-ball computing for scalable graph representation learning
Researchers have introduced SCGNN, a novel framework designed to enhance graph neural networks by improving the capture of semantic consistency among nodes. This approach utilizes granular-ball computing (GBC) to effici…
-
New framework uses causal modeling to advance edge classification in graphs
Researchers have introduced the Causal Edge Classification Framework (CECF), a novel approach to edge classification on graphs. This framework uniquely models edge features as a high-dimensional treatment, accounting fo…
-
New GNN framework enhances recommender systems with dynamic user similarity
Researchers have developed a new framework called DG-SA-GNN to improve recommender systems by incorporating dynamic user similarity graphs. This approach addresses limitations of traditional methods that rely on static …
-
Researchers use GNNs to analyze LLM-generated assurance cases
Researchers have developed a graph-based framework to analyze assurance cases, which are structured arguments used in regulated industries to justify system requirements and properties. This framework employs graph neur…
-
研究人员结合 DPU 和 GPU 以加速神经网络推理
研究人员开发了一种新颖的方法，通过在深度学习处理单元 (DPU) 和图形处理单元 (GPU) 之间拆分卷积神经网络 (CNN) 计算来加速神经网络推理。这种“拆分 CNN 推理”方法在数据源附近的 DPU 上处理初始层，在 GPU 上处理后续层，从而显著降低延迟。还引入了一个图神经网络 (GNN) 模型，以准确预测各种 CNN 架构的最佳层划分，准确率达到 96.27%。
-
Graph learning approach enhances SDN scalability for LEO mega-constellations
Researchers have developed a new software-defined networking (SDN) framework to manage the immense scale of Low Earth Orbit (LEO) satellite mega-constellations. This approach utilizes graph neural networks (GNNs) to mod…
-
综述梳理多智能体强化学习中的图神经网络
本文综述了利用图神经网络（GNN）进行通信的多智能体强化学习（MARL）的最新进展。文章重点介绍了GNN如何应用于交互图，使智能体能够共享信息并朝着共同目标进行更好的协调。作者旨在对MARL中这些基于GNN的通信方法进行结构化分类，使基本概念更易于理解。
-
GraphPL uses GNNs for robust modality imputation in patchwork learning
Researchers have introduced GraphPL, a novel approach for handling missing data in distributed multi-modal learning scenarios. This method utilizes graph neural networks to effectively impute incomplete modalities acros…
-
AI models Ligandformer and protein dynamics survey advance drug discovery and biological research
Researchers have developed Ligandformer, a Graph Neural Network designed to predict compound properties with enhanced interpretability. This model integrates attention maps to reveal how specific structural features inf…
-
New Graph Transformer models improve microservice tail latency prediction
Two new research papers propose advanced methods for predicting tail latency in microservice systems. The first, STLGT, uses a graph transformer to model service dependencies and a temporal module for workload dynamics,…
-
Hamiltonian Graph Inference Networks jointly discover structure and dynamics
Researchers have developed the Hamiltonian Graph Inference Network (HGIN), a novel method for simultaneously discovering the interaction structure and predicting the dynamics of lattice Hamiltonian systems from trajecto…
-
New graph-augmented segmentation enhances in situ inspection for 3D printing
Researchers have developed a novel graph-augmented segmentation method to improve in situ inspection of complex shapes in Laser Powder Bed Fusion (L-PBF) additive manufacturing. This approach utilizes a Graph Neural Net…