Movement Metric Learning - PLOS One

This repository contains the work published as a journal in PLOS One.

The purpose of this repository is to support reproducible research. It contains:

• the datasets
• the optimisation code
• the statistical analysis code
• the visualisation code

that were used to produce the graphs and the stastical analysis present in the manuscript.

You will need a python environement to run the code. For those using conda, it can be created with:

conda env create --file requirements.yaml

dataset

The raw files are provided as zip files under the release v1.0.0. Download and extract these under data.

optimisation code

The source code for all the computations presented in the paper is contained under app. In that directory, you can execute:

kedro run --env [exp] --pipeline [exp]

where [exp] is one of: exp1, spatial, temporal. This can be used to reproduce the baseline correlation grid search, the spatial optimisation and the temporal optimisation, respectively.

figure and analysis

Every figures and statistical analysis can be re-generated through a set of jupyter notebooks. These are located under figures.

Acknowledgements

Supported by the ELEMENT project (ANR-18-CE33-0002) and the ARCOL project (ANR-19-CE33-0001) from the French National Research Agency.