Official PyTorch Implementation for Advancing Bayesian Optimization via Learning Correlated Latent Space (CoBO) (arxiv).
Seunghun Lee*, Jaewon Chu*, Sihyeon Kim*, Juyeon Ko, Hyunwoo J. Kim, In Advanced in Neural Information Processing Systems (NeurIPS 2023).
![](https://private-user-images.githubusercontent.com/29230924/284174496-6cffa825-4c02-4f08-acd2-31e546c4458d.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.2yJWyY1nTW1xqUiuVLaLJ5Jy8x3wNpibZ2t9As8scYw)
We provide setup script file and environment file.
To setup the project, you can use the provided YAML file by running the following command:
conda env create -f requirements.yml
Or, for a shell script setup, run:
sh setup.sh
This repository uses tasks from the GuacaMol benchmark. Run a task with:
python3 scripts/molecule_optimization.py --task_id [TASK] run_cobo done
Available [TASK] codes include:
- med1: Median molecules 1
- pdop: Perindopril MPO
- adip: Amlodipine MPO
- rano: Ranolazine MPO
- osmb: Osimertinib MPO
- zale: Zaleplon MPO
- valt: Valsartan SMARTS
- med2: Median molecules 2
- siga: Sitagliptin MPO
- dhop: Deco Hop
- shop: Scaffold Hop
- fexo: Fexofenadine MPO
For more tasks, see the GuacaMol benchmark page.
@inproceedings{lee2023advancing,
title={Advancing Bayesian Optimization via Learning Correlated Latent Space},
author={Lee, Seunghun and Chu, Jaewon and Kim, Sihyeon and Ko, Juyeon and Kim, Hyunwoo J},
booktitle={Advances in Neural Information Processing Systems},
year={2023}
}
This repository is based on lolbo.
Code is released under MIT License.