AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

SEMIR: Semantic Minor-Induced Representation Learning on Graphs for Visual Segmentation

Luke James Miller, Yugyung Lee

ICML 2026 · Graph learning › Graph self-supervised learning · approach: Multi-view graph clustering · 0 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
add edge contraction (ablation)SEMIR · BraTS ETDice: 0.441 -> 0.894Improved
add edge deletion (ablation)SEMIR · BraTS ETDice: 0.719 -> 0.894Improved
add node deletion (ablation)SEMIR · BraTS ETDice: 0.812 -> 0.894Improved
learn minor parameters with few-shot optimization instead of manual tuning (ablation)SEMIR · BraTS ETDice: 0.837 -> 0.894Improved
learn minor parameters with few-shot optimization instead of manual tuning (ablation)SEMIR · NWPU VHR-10small-object IoU: 0.763 -> 0.789Improved
add relational edge features (ablation)SEMIR · BraTS ETDice: 0.725 -> 0.894Improved
add spatial features (compactness, elongation) (ablation)SEMIR · BraTS ETDice: 0.661 -> 0.894Improved
increase connectivity from 6 to 18 (ablation)SEMIR · NWPU VHR-10small-object IoU: +3-5% IoU at 2.5x runtimeMixed
SEMIR binary graph-minor model versus nnU-NetSEMIR · BraTS ET/TC, KiTS tumor, LiTS tumorDice: KiTS T 0.720 -> 0.819; LiTS T 0.733 -> 0.891Improved
SEMIR versus best published methods on aggregate regionsSEMIR · BraTS WT, LiTS LTDice: BraTS WT 0.943 (GTMamba) vs 0.920; LiTS LT 0.977 (Lgma-net) vs 0.959Hurt

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