Are Powerful Graph Neural Nets Necessary? A Dissection on Graph Classification

11 May 2019Ting ChenSong BianYizhou Sun

Graph Neural Nets (GNNs) have received increasing attentions, partially due to their superior performance in many node and graph classification tasks. However, there is a lack of understanding on what they are learning and how sophisticated the learned graph functions are... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Graph Classification COLLAB GFN-light Accuracy 81.34% # 4
Graph Classification COLLAB GFN Accuracy 81.50% # 3
Graph Classification D&D GFN Accuracy 78.78% # 8
Graph Classification D&D GFN-light Accuracy 78.62% # 9
Graph Classification ENZYMES GFN-light Accuracy 69.50% # 3
Graph Classification ENZYMES GFN Accuracy 70.17% # 2
Graph Classification IMDb-B GFN-light Accuracy 73.00% # 12
Graph Classification IMDb-B GFN Accuracy 73.00% # 12
Graph Classification IMDb-M GFN Accuracy 51.80% # 5
Graph Classification IMDb-M GFN-light Accuracy 51.20% # 7
Graph Classification MUTAG GFN-light Accuracy 89.89% # 11
Graph Classification MUTAG GFN Accuracy 90.84% # 8
Graph Classification NCI1 GFN-light Accuracy 81.43% # 14
Graph Classification NCI1 GFN Accuracy 83.65% # 9
Graph Classification PROTEINS GFN-light Accuracy 77.44% # 11
Graph Classification PROTEINS GFN Accuracy 76.46% # 17
Graph Classification RE-M12K GFN-light Accuracy 49.75% # 1
Graph Classification RE-M12K GFN Accuracy 49.43% # 2
Graph Classification RE-M5K GFN Accuracy 49.43% # 5
Graph Classification RE-M5K GFN-light Accuracy 49.75% # 4