AI Paper Atlas

Research map / Graph learning / GNN architectures

An Annotated Graph Model with Differential Degree Heterogeneity for Directed Networks

Stefan Stein, Chenlei Leng

JMLR 2023 · Graph learning › GNN architectures · approach: Oversmoothing and heterophily · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
select lambda by BIC instead of theory-based heuristic (ablation)Penalized MLE for annotated graph model · Simulation (small n)estimation error of alpha, beta, muImproved
select lambda by BIC instead of theory-based heuristic (ablation)Penalized MLE · Simulationestimation error of gammaNo clear effect
select lambda by theory-based heuristic instead of BIC (ablation)Penalized MLE · Simulation (growing n)probability of correct model selectionImproved
select lambda by theory-based heuristic instead of BIC (ablation)Penalized MLE · Simulation (small n)false negativesHurt
select lambda by BIC (ablation)Penalized MLE · Simulation (large n)probability of correct model selectionHurt
increase network size n from 200 to 800 (ablation)Penalized MLE confidence interval for gamma_0,1 · Simulationmedian CI length (heuristic): 0.265 -> 0.075Improved
use BIC versus heuristic lambda for gamma_0,1 inference (ablation)Penalized MLE · Simulationcoverage at n=200: 0.952 -> 0.962No clear effect
use sparse penalized model on all 71 nodes instead of dense MLE model on 63 nodesAnnotated graph model · Lazega lawyer networkstandard errors: 0.10 vs 0.16 (Same status SE)Improved
use annotated graph model with sparse alpha/beta instead of Yan et al. modelAnnotated graph model vs Yan et al. · Lazega lawyer networkSame law school coefficient sign: 0.29 vs -0.48Mixed
use annotated graph model instead of Yan et al. modelSina Weibo network · Sina Weibo networkcoefficient signs without bias correction: -0.127 vs -0.391 (posts)Improved
vanilla 10-fold cross-validation link predictionAnnotated graph model · Lazega lawyer network link predictionAUC: high 80% rangeImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy