TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Node Classification	Amazon Computers	CoLinkDist	Accuracy	89.42%	# 6
Node Classification	Amazon Computers	CoLinkDistMLP	Accuracy	88.85%	# 7
Node Classification	Amazon Computers	LinkDist	Accuracy	89.49%	# 4
Node Classification	Amazon Computers	LinkDistMLP	Accuracy	89.44%	# 5
Node Classification	Amazon Photo	LinkDistMLP	Accuracy	93.83%	# 4
Node Classification	Amazon Photo	CoLinkDist	Accuracy	94.36%	# 2
Node Classification	Amazon Photo	CoLinkDistMLP	Accuracy	94.12%	# 3
Node Classification	Amazon Photo	LinkDist	Accuracy	93.75%	# 5
Node Classification	Citeseer	CoLinkDistMLP	Accuracy	75.77%	# 20
Node Classification	Citeseer	LinkDist	Accuracy	74.72%	# 26
Node Classification	Citeseer	CoLinkDist	Accuracy	75.79%	# 19
Node Classification	Citeseer	LinkDistMLP	Accuracy	75.25%	# 24
Node Classification	CiteSeer with Public Split: fixed 20 nodes per class	LinkDistMLP	Accuracy	70.26%	# 31
Node Classification	CiteSeer with Public Split: fixed 20 nodes per class	CoLinkDistMLP	Accuracy	70.96%	# 28
Node Classification	CiteSeer with Public Split: fixed 20 nodes per class	CoLinkDist	Accuracy	70.79%	# 29
Node Classification	CiteSeer with Public Split: fixed 20 nodes per class	LinkDist	Accuracy	70.27%	# 30
Node Classification	Coauthor CS	CoLinkDist	Accuracy	95.80%	# 2
Node Classification	Coauthor CS	LinkDist	Accuracy	95.66%	# 5
Node Classification	Coauthor CS	CoLinkDistMLP	Accuracy	95.74%	# 3
Node Classification	Coauthor CS	LinkDistMLP	Accuracy	95.68%	# 4
Node Classification	Coauthor Physics	LinkDist	Accuracy	96.87%	# 5
Node Classification	Coauthor Physics	LinkDistMLP	Accuracy	96.91%	# 3
Node Classification	Coauthor Physics	CoLinkDist	Accuracy	97.05%	# 2
Node Classification	Coauthor Physics	CoLinkDistMLP	Accuracy	96.87%	# 5
Node Classification	Cora	LinkDistMLP	Accuracy	87.58%	# 15
Node Classification	Cora	LinkDist	Accuracy	88.24%	# 11
Node Classification	Cora	CoLinkDist	Accuracy	87.89%	# 13
Node Classification	Cora	CoLinkDistMLP	Accuracy	87.54%	# 16
Node Classification	Cora Full	LinkDistMLP	Accuracy	69.53%	# 4
Node Classification	Cora Full	LinkDist	Accuracy	69.87%	# 2
Node Classification	Cora Full	CoLinkDist	Accuracy	70.32%	# 1
Node Classification	Cora Full	CoLinkDistMLP	Accuracy	69.83%	# 3
Node Classification	Cora Full with Public Split	CoLinkDist	Accuracy	57.05%	# 1
Node Classification	Cora Full with Public Split	LinkDist	Accuracy	55.87%	# 2
Node Classification	Cora Full with Public Split	CoLinkDistMLP	Accuracy	53.43%	# 3
Node Classification	Cora Full with Public Split	LinkDistMLP	Accuracy	51.78%	# 4
Node Classification	Cora with Public Split: fixed 20 nodes per class	CoLinkDistMLP	Accuracy	81.19%	# 25
Node Classification	Cora with Public Split: fixed 20 nodes per class	LinkDist	Accuracy	81.05%	# 26
Node Classification	Cora with Public Split: fixed 20 nodes per class	LinkDistMLP	Accuracy	80.79%	# 27
Node Classification	Cora with Public Split: fixed 20 nodes per class	CoLinkDist	Accuracy	81.39%	# 24
Node Property Prediction	ogbn-arxiv	CoLinkDistMLP	Test Accuracy	0.5638 ± 0.0016	# 77
Node Property Prediction	ogbn-arxiv	CoLinkDistMLP	Validation Accuracy	0.5807 ± 0.0011	# 75
Node Property Prediction	ogbn-arxiv	CoLinkDistMLP	Number of params	120912	# 64
Node Property Prediction	ogbn-arxiv	CoLinkDistMLP	Ext. data	No	# 1
Node Property Prediction	ogbn-mag	CoLinkDistMLP	Test Accuracy	0.2761 ± 0.0018	# 36
Node Property Prediction	ogbn-mag	CoLinkDistMLP	Validation Accuracy	0.2646 ± 0.0013	# 36
Node Property Prediction	ogbn-mag	CoLinkDistMLP	Number of params	278202	# 36
Node Property Prediction	ogbn-mag	CoLinkDistMLP	Ext. data	No	# 1
Node Property Prediction	ogbn-products	CoLinkDistMLP	Test Accuracy	0.6259 ± 0.0010	# 60
Node Property Prediction	ogbn-products	CoLinkDistMLP	Validation Accuracy	0.7721 ± 0.0015	# 56
Node Property Prediction	ogbn-products	CoLinkDistMLP	Number of params	115806	# 50
Node Property Prediction	ogbn-products	CoLinkDistMLP	Ext. data	No	# 1
Node Classification	Pubmed	CoLinkDistMLP	Accuracy	89.53%	# 10
Node Classification	Pubmed	LinkDist	Accuracy	88.86%	# 14
Node Classification	Pubmed	LinkDistMLP	Accuracy	88.79%	# 15
Node Classification	Pubmed	CoLinkDist	Accuracy	89.58%	# 9
Node Classification	PubMed with Public Split: fixed 20 nodes per class	LinkDistMLP	Accuracy	72.41%	# 35
Node Classification	PubMed with Public Split: fixed 20 nodes per class	CoLinkDistMLP	Accuracy	75.41%	# 33
Node Classification	PubMed with Public Split: fixed 20 nodes per class	CoLinkDist	Accuracy	75.64%	# 32
Node Classification	PubMed with Public Split: fixed 20 nodes per class	LinkDist	Accuracy	74.06%	# 34

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-cora-full)](https://paperswithcode.com/sota/node-classification-on-cora-full?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-cora-full-with-public)](https://paperswithcode.com/sota/node-classification-on-cora-full-with-public?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-amazon-photo-1)](https://paperswithcode.com/sota/node-classification-on-amazon-photo-1?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-coauthor-cs)](https://paperswithcode.com/sota/node-classification-on-coauthor-cs?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-coauthor-physics)](https://paperswithcode.com/sota/node-classification-on-coauthor-physics?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-amazon-computers-1)](https://paperswithcode.com/sota/node-classification-on-amazon-computers-1?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-pubmed)](https://paperswithcode.com/sota/node-classification-on-pubmed?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-cora)](https://paperswithcode.com/sota/node-classification-on-cora?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-citeseer)](https://paperswithcode.com/sota/node-classification-on-citeseer?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-cora-with-public-split)](https://paperswithcode.com/sota/node-classification-on-cora-with-public-split?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-citeseer-with-public)](https://paperswithcode.com/sota/node-classification-on-citeseer-with-public?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-classification-on-pubmed-with-public)](https://paperswithcode.com/sota/node-classification-on-pubmed-with-public?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-property-prediction-on-ogbn-mag)](https://paperswithcode.com/sota/node-property-prediction-on-ogbn-mag?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-property-prediction-on-ogbn-products)](https://paperswithcode.com/sota/node-property-prediction-on-ogbn-products?p=distilling-self-knowledge-from-contrastive)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/distilling-self-knowledge-from-contrastive/node-property-prediction-on-ogbn-arxiv)](https://paperswithcode.com/sota/node-property-prediction-on-ogbn-arxiv?p=distilling-self-knowledge-from-contrastive)`

Distilling Self-Knowledge From Contrastive Links to Classify Graph Nodes Without Passing Messages

16 Jun 2021 · Yi Luo, Aiguo Chen, Ke Yan, Ling Tian ·

Nowadays, Graph Neural Networks (GNNs) following the Message Passing paradigm become the dominant way to learn on graphic data. Models in this paradigm have to spend extra space to look up adjacent nodes with adjacency matrices and extra time to aggregate multiple messages from adjacent nodes. To address this issue, we develop a method called LinkDist that distils self-knowledge from connected node pairs into a Multi-Layer Perceptron (MLP) without the need to aggregate messages. Experiment with 8 real-world datasets shows the MLP derived from LinkDist can predict the label of a node without knowing its adjacencies but achieve comparable accuracy against GNNs in the contexts of semi- and full-supervised node classification. Moreover, LinkDist benefits from its Non-Message Passing paradigm that we can also distil self-knowledge from arbitrarily sampled node pairs in a contrastive way to further boost the performance of LinkDist.

PDF Abstract

Code

Add Remove Mark official

cf020031308/LinkDist official

cf020031308/LinkDist

Tasks

Add Remove

Node Classification

Node Property Prediction

Datasets

Pubmed

OGB

Cora Citeseer PPI Amazon Photo

Results from the Paper

Edit

Ranked #1 on Node Classification on Cora Full

Get a GitHub badge

Task	Dataset	Model	Metric Name	Metric Value	Global Rank	Benchmark
Node Classification	Amazon Computers	CoLinkDist	Accuracy	89.42%	# 6	Compare
Node Classification	Amazon Computers	CoLinkDistMLP	Accuracy	88.85%	# 7	Compare
Node Classification	Amazon Computers	LinkDist	Accuracy	89.49%	# 4	Compare
Node Classification	Amazon Computers	LinkDistMLP	Accuracy	89.44%	# 5	Compare
Node Classification	Amazon Photo	LinkDistMLP	Accuracy	93.83%	# 4	Compare
Node Classification	Amazon Photo	CoLinkDist	Accuracy	94.36%	# 2	Compare
Node Classification	Amazon Photo	CoLinkDistMLP	Accuracy	94.12%	# 3	Compare
Node Classification	Amazon Photo	LinkDist	Accuracy	93.75%	# 5	Compare
Node Classification	Citeseer	CoLinkDistMLP	Accuracy	75.77%	# 20	Compare
Node Classification	Citeseer	LinkDist	Accuracy	74.72%	# 26	Compare
Node Classification	Citeseer	CoLinkDist	Accuracy	75.79%	# 19	Compare
Node Classification	Citeseer	LinkDistMLP	Accuracy	75.25%	# 24	Compare
Node Classification	CiteSeer with Public Split: fixed 20 nodes per class	LinkDistMLP	Accuracy	70.26%	# 31	Compare
Node Classification	CiteSeer with Public Split: fixed 20 nodes per class	CoLinkDistMLP	Accuracy	70.96%	# 28	Compare
Node Classification	CiteSeer with Public Split: fixed 20 nodes per class	CoLinkDist	Accuracy	70.79%	# 29	Compare
Node Classification	CiteSeer with Public Split: fixed 20 nodes per class	LinkDist	Accuracy	70.27%	# 30	Compare
Node Classification	Coauthor CS	CoLinkDist	Accuracy	95.80%	# 2	Compare
Node Classification	Coauthor CS	LinkDist	Accuracy	95.66%	# 5	Compare
Node Classification	Coauthor CS	CoLinkDistMLP	Accuracy	95.74%	# 3	Compare
Node Classification	Coauthor CS	LinkDistMLP	Accuracy	95.68%	# 4	Compare
Node Classification	Coauthor Physics	LinkDist	Accuracy	96.87%	# 5	Compare
Node Classification	Coauthor Physics	LinkDistMLP	Accuracy	96.91%	# 3	Compare
Node Classification	Coauthor Physics	CoLinkDist	Accuracy	97.05%	# 2	Compare
Node Classification	Coauthor Physics	CoLinkDistMLP	Accuracy	96.87%	# 5	Compare
Node Classification	Cora	LinkDistMLP	Accuracy	87.58%	# 15	Compare
Node Classification	Cora	LinkDist	Accuracy	88.24%	# 11	Compare
Node Classification	Cora	CoLinkDist	Accuracy	87.89%	# 13	Compare
Node Classification	Cora	CoLinkDistMLP	Accuracy	87.54%	# 16	Compare
Node Classification	Cora Full	LinkDistMLP	Accuracy	69.53%	# 4	Compare
Node Classification	Cora Full	LinkDist	Accuracy	69.87%	# 2	Compare
Node Classification	Cora Full	CoLinkDist	Accuracy	70.32%	# 1	Compare
Node Classification	Cora Full	CoLinkDistMLP	Accuracy	69.83%	# 3	Compare
Node Classification	Cora Full with Public Split	CoLinkDist	Accuracy	57.05%	# 1	Compare
Node Classification	Cora Full with Public Split	LinkDist	Accuracy	55.87%	# 2	Compare
Node Classification	Cora Full with Public Split	CoLinkDistMLP	Accuracy	53.43%	# 3	Compare
Node Classification	Cora Full with Public Split	LinkDistMLP	Accuracy	51.78%	# 4	Compare
Node Classification	Cora with Public Split: fixed 20 nodes per class	CoLinkDistMLP	Accuracy	81.19%	# 25	Compare
Node Classification	Cora with Public Split: fixed 20 nodes per class	LinkDist	Accuracy	81.05%	# 26	Compare
Node Classification	Cora with Public Split: fixed 20 nodes per class	LinkDistMLP	Accuracy	80.79%	# 27	Compare
Node Classification	Cora with Public Split: fixed 20 nodes per class	CoLinkDist	Accuracy	81.39%	# 24	Compare
Node Property Prediction	ogbn-arxiv	CoLinkDistMLP	Test Accuracy	0.5638 ± 0.0016	# 77	Compare
			Validation Accuracy	0.5807 ± 0.0011	# 75	Compare
			Number of params	120912	# 64	Compare
			Ext. data	No	# 1	Compare
Node Property Prediction	ogbn-mag	CoLinkDistMLP	Test Accuracy	0.2761 ± 0.0018	# 36	Compare
			Validation Accuracy	0.2646 ± 0.0013	# 36	Compare
			Number of params	278202	# 36	Compare
			Ext. data	No	# 1	Compare
Node Property Prediction	ogbn-products	CoLinkDistMLP	Test Accuracy	0.6259 ± 0.0010	# 60	Compare
			Validation Accuracy	0.7721 ± 0.0015	# 56	Compare
			Number of params	115806	# 50	Compare
			Ext. data	No	# 1	Compare
Node Classification	Pubmed	CoLinkDistMLP	Accuracy	89.53%	# 10	Compare
Node Classification	Pubmed	LinkDist	Accuracy	88.86%	# 14	Compare
Node Classification	Pubmed	LinkDistMLP	Accuracy	88.79%	# 15	Compare
Node Classification	Pubmed	CoLinkDist	Accuracy	89.58%	# 9	Compare
Node Classification	PubMed with Public Split: fixed 20 nodes per class	LinkDistMLP	Accuracy	72.41%	# 35	Compare
Node Classification	PubMed with Public Split: fixed 20 nodes per class	CoLinkDistMLP	Accuracy	75.41%	# 33	Compare
Node Classification	PubMed with Public Split: fixed 20 nodes per class	CoLinkDist	Accuracy	75.64%	# 32	Compare
Node Classification	PubMed with Public Split: fixed 20 nodes per class	LinkDist	Accuracy	74.06%	# 34	Compare

Methods

Add Remove

No methods listed for this paper. Add relevant methods here

Edit Social Preview

Distilling Self-Knowledge From Contrastive Links to Classify Graph Nodes Without Passing Messages

Code Edit Add Remove Mark official

Tasks Edit Add Remove

Datasets Edit

Results from the Paper Edit

Methods Edit Add Remove

Code

Add Remove Mark official

Tasks

Add Remove

Datasets

Results from the Paper

Edit

Methods

Add Remove