Polymer Informatics with Multi-Task Learning

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