AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

Analyzing Data-Centric Properties for Graph Contrastive Learning

Puja Trivedi, Ekdeep Singh Lubana, Mark Heimann, Danai Koutra, Jayaraman J. Thiagarajan

NeurIPS 2022 · Graph learning › Graph self-supervised learning · approach: Contrastive learning with negatives and views · 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
train graph SSL methods with generic graph augmentations vs untrained encoderGraphCL, GAE, AAGAE, SpecCL, SimSiam, BYOL · 7 graph benchmarks (Reddit etc.)kNN accuracyNo clear effect
increase invariance to generic augmentationsGraphCL and other graph SSL methods · RedditkNN accuracyNo clear effect
use content-aware edge dropping instead of random edge droppingGraphCL · Synthetic, increasing style ratiolinear probe accuracyImproved
use content-aware edge dropping instead of random edge droppingSpecLoss · Synthetic motif+tree graph datasetlinear probe accuracyImproved
use content-aware augmentationsGAE and AAGAE reconstruction methods · Synthetic motif+tree graph datasetlinear probe accuracyNo clear effect
use learned augmentor (AD-GCL)GraphCL-style contrastive learning · Synthetic motif+tree graph datasetlinear probe accuracyHurt
add content-preserving edge dropping to JOAO augmentation priorJOAO · Synthetic motif+tree graph datasetlinear probe accuracyNo clear effect
use implicit weight-space perturbation (SimGRACE)SimGRACE · Synthetic motif+tree graph datasetlinear probe accuracyHurt
use content-preserving instead of random edge dropping at 20% strengthGraphCL · Synthetic, kappa=6separabilityImproved
use content-preserving instead of random edge dropping at 20% strengthGraphCL · Synthetic, kappa=6invarianceNo clear effect
raise generic augmentation strength to 60%GraphCL · Synthetic, kappa=6invariance and separabilityHurt
use biochemistry-inspired content-aware augmentations instead of mask edge featuresGraphCL-style SSL · BACEinvariance and separabilityImproved

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