graph neural networks
PulseAugur coverage of graph neural networks — every cluster mentioning graph neural networks across labs, papers, and developer communities, ranked by signal.
- 2026-05-25 research_milestone Researchers proposed new polynomial-time algorithms for explaining Graph Neural Networks. 来源
- 2026-05-13 research_milestone A new graph neural network architecture was introduced for the multicut problem. 来源
- 2026-05-11 research_milestone A new method for pre-training GNNs using ECFPs shows improved performance in QSAR tasks. 来源
13 天有情绪数据
-
GP2F method enhances cross-domain graph neural network adaptation
Researchers have introduced GP2F, a novel method for cross-domain graph prompting that aims to improve the adaptation of pre-trained graph neural networks to new tasks. The method is based on theoretical analysis showin…
-
New logic-based graph learning method rivals GNNs in speed and performance
Researchers have developed new variants of the Weisfeiler-Leman algorithm for graph classification, which involve modifying the underlying logical framework. These variants allow graph data to be tabularized, enabling t…
-
New GNN defense uses self-supervised purifier against adversarial attacks
Researchers have developed a novel self-supervised adversarial purification framework for Graph Neural Networks (GNNs). This new method separates the task of robustness from classification by using a dedicated purifier,…
-
GNNs enhance physics simulations by learning model discrepancies
Researchers have developed a novel hybrid twin framework that combines physics-based models with Graph Neural Networks (GNNs) to improve simulations of complex physical phenomena. This approach addresses the limitations…
-
GILT model offers LLM-free, tuning-free graph learning
Researchers have introduced GILT, a novel graph foundational model designed to overcome limitations in handling heterogeneous graph data. Unlike existing models that rely on Large Language Models or require extensive pe…
-
New algorithms improve GNN explainability by reducing walk search complexity
Researchers have developed new polynomial-time algorithms to address the exponential computational complexity of identifying relevant walks in Graph Neural Networks (GNNs). This advancement significantly improves the ap…
-
New PRiSM method offers complete graph canonicalization for GNNs
Researchers have demonstrated that the Weisfeiler-Leman (WL) test, a common method for graph isomorphism testing, is incomplete for graphs with simple spectra. This limitation extends to Graph Neural Networks (GNNs) tha…
-
Graph condensation methods need reset, paper argues
A new position paper argues that the current methods for graph condensation, a technique aimed at making Graph Neural Networks (GNNs) more scalable, are fundamentally flawed. The paper highlights that existing approache…
-
New framework GraphSSR improves LLM-based zero-shot graph learning
Researchers have developed GraphSSR, a new framework to improve zero-shot graph learning by adaptively extracting and denoising subgraphs. This approach addresses the limitations of current methods that use a one-size-f…
-
New GNN architecture learns adaptive graph geometry for better long-range task performance
Researchers have developed a novel Graph Neural Network (GNN) architecture called mu-ChebNet, designed to improve performance on long-range graph tasks. This architecture learns a node-wise weighting function that modif…
-
GNN explanation methods reveal disease signatures in biological networks
Researchers have evaluated four popular explanation methods for graph neural networks (GNNs) to understand their effectiveness in identifying disease-associated structures within biological networks. Using synthetic dat…
-
New algorithms efficiently explain graph neural network decisions
Researchers have developed new algorithms to efficiently explain the decision-making processes of graph neural networks (GNNs). These methods, based on message passing techniques, significantly reduce the computational …
-
New method Ex-GraphRAG deciphers LLM evidence routing from knowledge graphs
Researchers have developed Ex-GraphRAG, a novel method for interpreting how Large Language Models (LLMs) use information from knowledge graphs. This new approach replaces the standard Graph Neural Network encoder with a…
-
Deep ensembles fail to capture uncertainty in graph neural networks
A new research paper questions the effectiveness of deep ensembles for uncertainty quantification in graph neural networks. The study found that ensembles offer minimal improvement over single models, with gains primari…
-
Simple Random Node Sampling outperforms full-graph training for GNNs
Researchers have found that a simple Random Node Sampling (RNS) method for training Graph Neural Networks (GNNs) can match or exceed the performance of full-graph training. This surprising result holds true across numer…
-
Hybrid physics-informed neural networks advance electricity system design
A new review paper explores the use of hybrid physics-informed neural networks (PIML) for enhancing electricity systems. These methods embed physical laws into machine learning models, improving accuracy and efficiency,…
-
FROG framework learns relational database graph structures for deep learning
Researchers have developed FROG, a novel framework for Relational Deep Learning (RDL) that addresses the limitations of fixed graph structures in modeling relational databases. FROG introduces a learnable approach to gr…
-
Gaussian Sheaf Neural Networks leverage sheaf theory for Gaussian data
Researchers have introduced Gaussian Sheaf Neural Networks (GSNNs), a novel framework designed for learning on relational data where node features are represented by probability distributions, specifically Gaussian dist…
-
Graph Navier Stokes Networks tackle oversmoothing with convection
Researchers have introduced Graph Navier Stokes Networks (GNSN), a new architecture designed to address the oversmoothing problem in Graph Neural Networks. Unlike traditional diffusion-based methods, GNSN incorporates c…
-
New B-cos GNNs offer faster, inherent model explainability
Researchers have developed B-cos GNNs, a new type of graph neural network designed for inherent explainability. These models decompose predictions into per-node, per-feature contributions using a dynamic linearity, elim…