
Large language models can enhance graph structure learning by denoising noisy connections and uncovering implicit node-wise dependencies.


coremsg

graphedit-large-language-models-for-graph-structure-learning


GraphEdit: Large Language Models for Graph Structure Learning



Large Language Models (LLMs) enhance graph quality for Multivariate Time-Series (MTS) data representation learning.



K-Link: Knowledge-Link Graph from LLMs for Enhanced Representation Learning in Multivariate Time-Series Data



Local Message Compensation (LMC) is a subgraph-wise sampling method with provable convergence, accelerating training efficiency for GNNs.



LMC: Fast Training of GNNs via Subgraph-wise Sampling with Provable Convergence at ICLR 2023



Standard-GNNs übertreffen spezialisierte Modelle bei der Bewertung unter Heterophilie.



Eine kritische Betrachtung der Bewertung von GNNs unter Heterophilie
