Predicting molecular properties using message passing neural networks (MPNN). This repository contains some of my code from my work on the Kaggle "Predicting Molecular Properties" Challenge in 2019. My code features two implementations of MPNN models (PyTorch and Tensorflow (Deepchem)).
MPNNs are a recent class of neural networks suitable for graph data and has been used to perform learning of molecular structure. A basic introduction to MPNNs can be found here.