- Ref paper: Automatic chemical design using a data-driven continuous representation of molecules - 2018
- Epoch vs. Train loss
- Note 1: Generated SMILES from this model is tested with RDkit library to know whether or not it is a valid SMILES.
I observed that the percentage of valid generated SMILES from this model is very low (<0.5%)
A good generated SMILES for dug discovery is more important than the percentage of valid generated SMILES,
but the model itself cannot guarantee that the valid generated SMILES is a good candidate.
It indicated the limitation of the model for using in production. - Note 2: In the next repository, the VAE model combined with Reinforcement Learning will be implemented and
hopefully, it can generate SMILES better than this model.
-
Notifications
You must be signed in to change notification settings - Fork 1
tuantla80/VAE-Molecular-Generation
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
Molecular generative model based on Variational AutoEncoder (VAE)
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published