OrbNet: Deep Learning for Quantum Chemistry Using Symmetry-Adapted Atomic-Orbital Features
Proposal
Main paper can be found here.
- ML based method to predict energy solutions from the Schroedinger equation
- Use symmetry adapted atomic orbitals + graph neural networks as their architecture for this purpose
- This architecture is called as OrbNet
Summary
- This is a regression model that is being learnt through DL techniques
- Regression targets themselves are quantum mechanical energies
- the GNN model will have edge and node attention built into it