Read the original document by opening this link in a new tab.
Table of Contents
1. Introduction
2. Message Passing Neural Networks
3. Related Work
Summary
Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. In this paper, the authors reformulate existing models into a common framework called Message Passing Neural Networks (MPNNs) and explore novel variations within this framework. They demonstrate state-of-the-art results on molecular property prediction benchmarks and emphasize the importance of applying machine learning methods to chemistry problems. The paper discusses various models from the literature that can be described using the MPNN framework and highlights the need for careful empirical studies to improve these models. The authors also address the challenges in quantum chemistry calculations and the limitations of existing methods. Overall, the paper presents a significant step towards making well-designed MPNNs the standard for supervised learning on molecules.