Expressivity and Generalization: Fragment-Biases for Molecular GNNs

Keyphrases

• Graph Neural Network 100%
• Molecular Graph 100%
• Weisfeiler-Lehman 50%
• Low Error 33%
• Expressiveness 33%
• Recent Advances 16%
• Neural Network Architecture 16%
• Generalization Ability 16%
• Transformer Architecture 16%
• Molecular Modeling 16%
• Empirical Performance 16%
• Performance Prediction 16%
• Modelling Task 16%
• Robust Solution 16%
• Theoretic Analysis 16%
• Molecular Properties 16%
• Inductive Bias 16%
• Biased Models 16%
• Biased Graph 16%
• Higher-rank Graph 16%

Computer Science

• Graph Neural Network 100%
• Neural Network Architecture 16%
• Theoretic Analysis 16%
• Robust Solution 16%
• Predictive Performance 16%