Polymer Informatics with Multi-Task Learning

IMPORTANT NOTE: The code and data shared here is available for academic non-commercial use only

This repository contains the code to the paper. The code trains four different machine learning models for the prediction of polymer properties. Two of the models are single-task (nn_st.py and gpr.py) and two are multi-task (nn_mt2.py and nn_mt.py) models. Please see the paper for more details.

Prerequisites

• Poetry must be installed. See https://python-poetry.org/docs/#installation

Install

1. Clone repo

git clone https://gitlab.com/ramprasad-group/multi-task-learning && cd multi-task-learning

2. Init poetry

poetry install
poetry shell
cd .. && mkdir test && cd test

3. Run

mtask -h

Use

1. In the test directory

mkdir dataset && cd dataset

2. Place dataset in the dataset/ directory

3. Run

mtask_new nn-st train
or
mtask_new nn-mt train
or
mtask_new nn-st2 train