ptc-cmc / imodels-supplements

iMoDELS supplemental information

This repository is intended to host the supplemental information for the paper title: "High-throughput Screening of Tribological Properties of Monolayer Films using Molecular Dynamics and Machine Learning"

MD simulations data

The tribological data calculated from MD simulations can be found as csv files located here. These properties are calculated using the Amonton's Law of Friction. The friction force of normal load is collected from the last 5 ns of the MD production run. Please refer to the main paper for more details.

Machine Learning (Random Forest Regressor)

This study utilized the Random Forest Regressor provided by scikit-learn. Multiple ML models were created, each varied by the number data points of their training set and/or the seed used to select data points out of the main data set.

Using this repository

To set up the environment neccessary for this repo

    conda env create -f env.yml
    conda activate screening

This repository includes codes needed to: 1. Train and pickle the ML models: trainML.py 2. Evaluate the created ML models by applying them to common test set: predictML.py 3. Analyze/visualize the final result: series of iPython notebook located here 4. Once the pickled ML models have been created, the Data-Lookup.ipynb can be used to look up data from the main data set, or utilized the created ML models to estimate tribological properties of any systems.