Ranking Structured Objects with Graph Neural Networks

18 Apr 2021  ·  Clemens Damke, Eyke Hüllermeier ·

Graph neural networks (GNNs) have been successfully applied in many structured data domains, with applications ranging from molecular property prediction to the analysis of social networks. Motivated by the broad applicability of GNNs, we propose the family of so-called RankGNNs, a combination of neural Learning to Rank (LtR) methods and GNNs. RankGNNs are trained with a set of pair-wise preferences between graphs, suggesting that one of them is preferred over the other. One practical application of this problem is drug screening, where an expert wants to find the most promising molecules in a large collection of drug candidates. We empirically demonstrate that our proposed pair-wise RankGNN approach either significantly outperforms or at least matches the ranking performance of the naive point-wise baseline approach, in which the LtR problem is solved via GNN-based graph regression.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Graph Ranking ogbg-molesol 2-WL-GNN + CmpNN Kendall's Tau 0.718 # 1
Graph Ranking ogbg-molesol 2-WL-GNN + DirectRanker Kendall's Tau 0.745 # 3
Graph Ranking ogbg-molesol 2-WL-GNN + Utility Regression Kendall's Tau 0.747 # 4
Graph Ranking ogbg-molesol 2-WL-GNN + Rank Regression Kendall's Tau 0.720 # 2
Graph Ranking ogbg-molfreesolv 2-WL-GNN + Utility Regression Kendall's Tau 0.379 # 1
Graph Ranking ogbg-molfreesolv 2-WL-GNN + CmpNN Kendall's Tau 0.527 # 4
Graph Ranking ogbg-molfreesolv 2-WL-GNN + DirectRanker Kendall's Tau 0.525 # 3
Graph Ranking ogbg-molfreesolv 2-WL-GNN + Rank Regression Kendall's Tau 0.524 # 2
Graph Ranking ogbg-mollipo 2-WL-GNN + Utility Regression Kendall's Tau 0.318 # 1
Graph Ranking ogbg-mollipo 2-WL-GNN + CmpNN Kendall's Tau 0.503 # 3
Graph Ranking ogbg-mollipo 2-WL-GNN + DirectRanker Kendall's Tau 0.505 # 4
Graph Ranking ogbg-mollipo 2-WL-GNN + Rank Regression Kendall's Tau 0.332 # 2
Graph Ranking ZINC 2-WL-GNN + Rank Regression Kendall's Tau 0.810 # 3
Graph Ranking ZINC 2-WL-GNN + CmpNN Kendall's Tau 0.873 # 2
Graph Ranking ZINC 2-WL-GNN + Utility Regression Kendall's Tau 0.803 # 4
Graph Ranking ZINC 2-WL-GNN + DirectRanker Kendall's Tau 0.894 # 1

Methods


No methods listed for this paper. Add relevant methods here