AI Paper Atlas

Research map / Graph learning / GNN architectures

Learning Network Dismantling Without Handcrafted Inputs

Haozhe Tian, Pietro Ferraro, Robert Noel Shorten, Mahdi Jalili, Homayoun Hamedmoghadam

AAAI 2026 · Graph learning › GNN architectures · approach: Expressive equivariant message passing · 2 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
use MIND without handcrafted input features instead of feature-based GNN dismantlers (GDM)MIND · Real networks, all domainsrelative dismantling AUC: 104.13 -> 100.0Improved
add all-to-one attention mechanism (MIND-AM) replacing softmax-normalized attention (ablation)MIND · Validation AUC on synthetic networksvalidation dismantling AUC: p = 8.1 × 10−4Improved
add message-iteration profiles (MIND-MP) over all message-passing iterations (ablation)MIND · Validation AUC on synthetic networksvalidation dismantling AUC: p = 4.5 × 10−2Improved
use GATv2 with softmax attention as message-passing operator with constant initial embeddings (ablation)MIND framework · Validation AUC on synthetic networksvalidation dismantling AUCHurt
use GCN as message-passing operator instead of MIND encoder (ablation)MIND framework · Validation AUC on synthetic networksvalidation dismantling AUCHurt
add degree-preserving edge rewiring to diversify training networks (ablation)MIND · Real networks, overalldismantling AUCImproved
add degree-preserving rewiring to training networks (ablation)MIND · roads-california (modularity 0.975)dismantling AUC: over 80% reduction in AUCImproved
add degree-preserving rewiring to training networks (ablation)MIND · munmun twitter social (assortativity -0.878)dismantling AUC: AUC reduced by 40%Improved
MIND vs exact methods on technological networksMIND · Technological networksdismantling AUCHurt
use handcrafted input features (GDM) instead of learned embeddingsGDM · Synthetic networks ER, CM, SBMdismantling AUCHurt
MIND vs MS on SBM 1k nodesMIND · Stochastic Block Model 1k nodesdismantling AUCNo 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