AI Paper Atlas

Research map / Graph learning / GNN architectures

Neural Dispersion on Graphs

Ryien Hosseini, Pouya Mahdi Gholami, Filippo Simini, Venkatram Vishwanath, Rebecca Willett, Henry Hoffmann

ICML 2026 · Graph learning › GNN architectures · approach: Theory of graph convolutional networks

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
use generator ensemble instead of single generator (ablation)NGD · Graph dispersion N=16GCD diversity: 6.100 -> 7.621Improved
use directional random-projection metrics instead of a single scalar projection (ablation)NGD · Graph dispersion N=16GCD diversity: 6.242 -> 7.621Improved
use generator ensemble instead of single generator (ablation)NGD · Graph dispersion N=128Portrait diversity: 0.929 -> 0.969Improved
NGD instead of ER-mix parametric ensembleNGD · Graph dispersion N=128Heat diversity: 0.068 -> 5.546Improved
NGD instead of combinatorial search at small NNGD · Graph dispersion N=16GCD diversity: 7.718 (Greedy->Genetic->Local) vs 7.621Mixed
NGD instead of combinatorial search at N=128NGD · Graph dispersion N=128Wave diversity: 10.974 (Greedy->Genetic) vs 10.674Hurt
scale N and k with NGDNGD · Large N up to 1024, k up to 8192diversityImproved
evaluate GIN on NGD graphs instead of generator-mix graphsGIN regressor · Spectral radius, N=256MAE: 3.2956 -> 10.5067Hurt
evaluate greedy coloring on NGD graphs instead of generator-mixgreedy coloring · Approximation ratio, largest-firstmean ratio: 0.25 -> 0.48Hurt

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