/linear-vs-blackbox

Code and data related to the publication: Interpretable models for extrapolation in scientific machine learning

Primary LanguageJupyter NotebookMIT LicenseMIT

Linear vs blackbox

This repository contains code and data for reproducing Interpretable models for extrapolation in scientific machine learning. No ongoing maintenance or support of this content should be expected.

Steps for reproduction

  1. download this folder and open a terminal inside it
  2. create a new python environment: python3 -m venv venv
  3. activate the python environment: . venv/bin/activate
  4. update pip in the python environment: python -m pip install -U pip
  5. install dependencies in the python environment: pip install -r requirements.txt
  6. start the jupyter server: jupyter lab
  7. open workflow.ipynb in jupyter and run it

Description of files

  • workflow.ipynb: juypter notebook which reproduces the paper
  • utils.py: python functions imported by workflow.ipynb
  • data
    • dataset_config.csv: specifies the name, target property, size, and source of each dataset
    • fig: figures generated by workflow.ipynb and used in the paper
    • raw: raw data files in CSV format

Distribution A Approved for Public Release, Distribution Unlimited