Baseline和SOTA实验结果汇总

ogbn-arxiv

Method Validation Acc(%) Test Acc(%)
MLP 57.65 ± 0.12 55.50 ± 0.23
Node2Vec 71.29 ± 0.13 70.07 ± 0.13
GCN 73.00 ± 0.17 71.74 ± 0.29
GraphSAGE 72.77 ± 0.16 71.49 ± 0.27
DeeperGCN 72.62 ± 0.14 71.92 ± 0.16
GCNII ---- 72.74 ± 0.16
JKNet(6) 73.35 ± 0.07 72.19 ± 0.21
GCN_res(SOTA,8) 73.69 ± 0.21 72.62 ± 0.37
GCN_res + FLAG(SOTA,8) 73.89 ± 0.12 72.76 ± 0.24

实验代码:

Cora、Citeseer、Pubmed

Test Acc(%)。

Method Cora Citeseer Pubmed
GCN 80.50 67.90 78.80
GAT 79.30 68.10 74.60
GraphSAGE 79.90 66.60 76.00
JKNet(6) 77.10 66.10 72.80
GCN_res(SOTA,64) 75.10 51.70 61.60

实验代码:

上一篇:AI 测试之Applitools入门教程


下一篇:An Improved Baseline for Sentence-level Relation Extraction