AI Paper Atlas

Research map / Graph learning / GNN architectures

The expressive power of pooling in Graph Neural Networks

Filippo Maria Bianchi, Veronica Lachi

NeurIPS 2023 · Graph learning › GNN architectures · approach: Subgraph and distance-aware GNNs · 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
replace non-expressive Top-k with expressive DMoN poolingGIN+pooling · EXPWL1test accuracy (%): 67.9 -> 99.0Improved
expressive pooling (DiffPool) instead of no poolingGIN · EXPWL1test accuracy (%): 99.3 -> 97.0No clear effect
replace Graclus-style regular pooling by Rand-dense random assignment (ablation)GIN+pooling · EXPWL1test accuracy (%): 99.9 -> 91.7Hurt
replace Graclus by Cmp-Graclus (complement graph) (ablation)GIN+pooling · EXPWL1test accuracy (%): 99.9 -> 91.9Hurt
random sparse scoring (Rand-sparse) instead of random dense (ablation)GIN+pooling · EXPWL1test accuracy (%): 91.7 -> 62.8Hurt
global_sum_pool instead of global_max_pool with 3 GIN layers (ablation)GIN · EXPWL1test accuracy (%): 98.3 -> 99.3Improved
global_sum_pool instead of global_max_pool with 2 GIN layers (ablation)GIN · EXPWL1test accuracy (%): 66.5 -> 92.1Improved
GIN layers instead of GCN layers, 2 layers sum pool (ablation)MP layers · EXPWL1test accuracy (%): 76.7 -> 92.1Improved
3 GIN layers instead of 2 GIN layers before global_max_pool (ablation)GIN · EXPWL1test accuracy (%): 66.5 -> 98.3Improved
increase pooling ratio from 0.05 to 0.2 for Top-k (ablation)Top-k · EXPWL1test accuracy (%): 69.7 -> 89.4Improved
increase pooling ratio from 0.05 to 0.2 for SAGPool (ablation)SAGPool · EXPWL1test accuracy (%): 61.6 -> 82.4Improved
increase pooling ratio from 0.05 to 0.2 for DMoN (expressive) (ablation)DMoN · EXPWL1test accuracy (%): 98.5 -> 98.1No clear effect

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