...
Stage | Milestone/Benchmark | Contributors | Deadline | Status | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
Curate training dataset | Work out best level of theory for the training dataset | November 310, 2024 |
| |||||||
Compute training dataset | December 31, 2024 |
| ||||||||
Curate testing dataset | Compile QM dataset | November 30, 2024 |
| |||||||
Compute QM dataset | January 31, 2025 |
| ||||||||
Compile simulation test set (Free Solv, maybe non-hydration solvation free energy sets that are harder to reproduce) | April 15, 2025 |
| ||||||||
Determine best NN architecture | Experiment with architecture, comparing GraphSAGE vs Attention-based GNNs with bond features | Implement attention-based GNN | December 31, 2024 |
| ||||||
Implement bond features in GraphSAGE (?) | December 31, 2024 |
| ||||||||
Determine best architecture | January 31, 2025 |
| ||||||||
First pass at NN training | Train using just ESPs, dipoles, quadrupoles, perhaps regularizing | Feb 28, 2025 |
| |||||||
Regularize to RESP charges or MBIS charges if buried atoms are a problem |
| |||||||||
Train directly to charge model if still having issues |
| |||||||||
Benchmark 1: QM | Neural network charge model with low testing error and equivalent or better performance to NAGL in simulationson QM data (ESPs, dipoles) | March 15, 2025 | ||||||||
Re-train VDW terms | March 30, 2025 |
| colour
| Green | ||||||
| title | Passed | |||||||||
| Status | ||||||||||
| colour | Red | title | Failed
| |||||||
Re-train valence terms | April 15, 2025 |
| ||||||||
Benchmark 2: Simulation | Neural network charge model with equivalent or better performance to NAGL in simulations | April 30, 2025 |
📊 Progress and findings
Curated data (or similar title)
...