This repository is the official implementation of How Non-Experts Perceive and Teach with Machine Learning Uncertainties. It contains:
benchmark
: jupyter notebooks presenting the benchmark presented in section 3.app
: the Marcelle source code of the application used in the user study presented in section 4.analysis
: the quantitative analysis of the user study logs. Results are presented in section 5.setup
: contains python configurations of the projectcommon
: contains common functions used both in the app, the benchmark and the analysis.
To install requirements in a conda environment: On MacOS:
pip install -r setup/conda_env_macos.yml
On Linux:
pip install -r setup/conda_env_linux.yml
To install the app:
cd app
npm install
or
cd app
yarn install
See the README in app
for more details.
To install Marcelle, please refers to the official website
The code here might be incomplete and not systematically documented. It aims at making the research process more transparent for reviewers and peers. Please contact me if you have any suggestions or find any problems.