Tune MPNN architecture
Open
ML
Priority: High
- Dominant language
- C++
- Stars
- 406
- Forks
- 134
- Avg merge
- 2d 11h
- Merged PRs (30d)
- 24
Description
Blocked by https://github.com/epam/Indigo/issues/659.
After we create a couple of useful presets of featurisers, it could be useful to try to improve MPNN architecture to achieve better predicting results.
**ToDo**
Tune various MPNN architecture params to achieve better predicting results on ADRA1A using out best preset of featurisers:
* Number of layers
* Number of message passing steps
* Activation function (tanh?)
Use best params as default.
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.