Home Console Docs News Papers Tutorials Datasets Wiki SOTA LLM Models GPU Leaderboard Events

English

Node Classification

Node Classification is a machine learning task in graph data analysis, aiming to predict unknown attributes of nodes in a graph based on the attributes of the nodes and their relationships. This task is implemented through models such as Graph Neural Networks, and its performance is typically evaluated using metrics like Accuracy and F1 on benchmark datasets such as Cora, Citeseer, and Pubmed. Node Classification has significant application value in areas like social network analysis, recommendation systems, and bioinformatics.

CiteSeer with Public Split: fixed 20 nodes per class

Chameleon (60%/20%/20% random splits)

Film (60%/20%/20% random splits)

Squirrel (60%/20%/20% random splits)

PubMed with Public Split: fixed 20 nodes per class

PubMed (60%/20%/20% random splits)

Texas (60%/20%/20% random splits)

Cora with Public Split: fixed 20 nodes per class

Cornell (60%/20%/20% random splits)

Wisconsin (60%/20%/20% random splits)

CiteSeer (60%/20%/20% random splits)

Cora (60%/20%/20% random splits)

Cora (48%/32%/20% fixed splits)

Citeseer (48%/32%/20% fixed splits)

PubMed (48%/32%/20% fixed splits)

Truncated Krylov

Coauthor Physics

JK (Heat Diffusion)

Truncated Krylov

Truncated Krylov

CiteSeer (0.5%)

Truncated Krylov

Amazon Computers

Cora Full-supervised

Cora: fixed 20 node per class

Self-supervised GraphMAE

GNNMoE(GCN-like P)

GCN+GAugM (Zhao et al., 2021)

Pubmed Full-supervised

GraphSAGE+DropEdge

Brazil Air-Traffic

Europe Air-Traffic

GCN_cheby (Kipf and Welling, 2017)

Citeseer Full-supervised

IncepGCN+DropEdge

USA Air-Traffic

DEMO-Net(weight)

DEMO-Net(weight)

ASDNet [ASDNet_ICCV2021]

Squirrel (48%/32%/20% fixed splits)

Chameleon (48%/32%/20% fixed splits)

Cora Full with Public Split

DEMO-Net(weight)

Cornell (48%/32%/20% fixed splits)

Chameleon(60%/20%/20% random splits)

Squirrel(60%/20%/20% random splits)

Cora: fixed 5 node per class

Texas (48%/32%/20% fixed splits)

Wisconsin (48%/32%/20% fixed splits)

Cora: fixed 10 node per class

Telegram (Directed Graph label rate 60%)

Citeseer: fixed 20 node per class

CiteSeer with Public Split: fixed 5 nodes per class

Film(48%/32%/20% fixed splits)

Pubmed random partition

Cora: fixed 20 nodes per class

Citeseer random partition

AMZ Computers: fixed 20 node per class

Pubmed: fixed 20 node per class

Cora random partition

CellTypeGraph Benchmark

plantcelltype-EdgeDeeperGCN