AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

Self-supervised Auxiliary Learning with Meta-paths for Heterogeneous Graphs

Dasol Hwang, Jinyoung Park, Sunyoung Kwon, KyungMin Kim, Jung-Woo Ha, Hyunwoo J Kim

NeurIPS 2020 · Graph learning › Graph self-supervised learning · approach: Semi-supervised node representation learning · 39 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 meta-path prediction auxiliary tasks with standard loss (ablation)GNNs · Last-FMAUC: +0.0106 avg gainImproved
learn auxiliary task weights via meta-learning (SELAR) instead of standard auxiliary loss (ablation)GNNs · Last-FMAUC: +0.0106 -> +0.0190 avg gainImproved
add Hint Network to SELAR (ablation)SELAR · Last-FMAUC: +0.0190 -> +0.0200 avg gainImproved
add Hint Network to SELAR (ablation)SELAR with GCN · Last-FMAUC: 0.8296 -> 0.8121Hurt
add meta-path auxiliary tasks and SELAR (ablation)GNNs · Book-CrossingAUC: +0.0263 avg gain (SELAR)Improved
add SELAR with meta-pathsGAT · IMDBF1: 0.5653 -> 0.6099Improved
add SELAR and Hint (ablation)GNNs · ACMF1: +0.0079 / +0.0090 avg gainImproved
learn sample weighting only without meta-path tasks (ablation)GNNs · Last-FMAUC: -0.0017 avg gainNo clear effect
learn sample weighting only without meta-path tasks (ablation)GNNs · IMDBF1: -0.0027 avg gainNo clear effect
use 3-fold meta cross-validation instead of 1-fold meta-data (ablation)SELAR · Last-FMAUC: 0.7885 -> 0.8296 (GCN)Improved

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