PMBO

Polymer Multiobjective Bayesian Optimization (PMBO) assiciated by physical descriptors

Description

Polymer Multiobjective Bayesian Optimization (PMBO) framework for exploring multifunctional polymers, including four components: 1) Polymer Library, which stores all polymer candidates; 2) Feature extraction, which based on physical feature engineering or graph descriptors generation; 3) Polymer properties simulator, which calculates polymer properties at different scales using DFT, MD and ML. 4) Multi-objective Bayesian algorithm, which predicts polymer properties and recommends potentially multifunctional polymers.Please refer to our work "Discovery of Multifunctional Polymers in Constrained Chemical Space Via Physical Descriptors-Guided Multi-objective Bayesian Optimization" for additional details.

Installation

Files loading and environment setup:

To download, clone this repository:

git clone https://github.com/SJTU-MI/PMBO.git

To run most code in this repository, the relevant anaconda environment can be installed from environment.yml. To build this environment, run：

cd ./PMBO
conda env create -f environment.yml
conda activate pmbo

Additionally, polymer physical feature engineering and properties calculations can be accessed at other GitHub repositories of APFEforPI and RadonPy.

Try the desired parts of the project:

Code in the pmbo folder

gp.py: Gaussian process regression model
Acq_fun.py: Acquisition functions such as EI(expected improvement) and UCB(upper confidence bound)
hypervolume.py: Calculation of hypervolume
utility.py: Utility functions such as Pareto front allocation and data pre/post processing
optimize.py: Core of multi-objective Bayesian optimization
log.py: PMBO Logo

Tutorial

MBO_tutorial.ipynb: A case of multi-objective optimization for multifunctional polymers discovery
example.csv:Benchmark dataset for testing (input file)
cal_data.csv: MBO recommended polymers and their observed properties (output file)
HV.csv: Optimized convergence curve evaluated by hypervolume (output file)

Authors

AUTHORS	Xiang Huang, Shenghong Ju
VERSION	V1.0 / July,2023
EMAILS	[email protected]
GROUP HOME	https://ju.sjtu.edu.cn/en/

Attribution

This work is under BSD-2-Clause License. Please, acknowledge use of this work with the appropiate citation to the repository and research article.