WORK IN PROGRESS
This is a Graph Convolutional Networks that aims to learn from molecular data represented as graphs. Ideally, the model would be able to encode a molecular structure and learn distributions from both specific entities in the graph (atoms and bonds in the molecule) as well as the overall structure (molecule). The model in interest is an objective reinforced generative model capable of learning from representation of inorganic molecules as well as viral structures.
While the base prototype is in working progress, data augmentation of molecules, while volatile, may be a future direction, as is the use of transfer learning to improve model performance.
Upwards and onwards, always and only 🚀!