First we describe the contents of the archive, then we suggest how to inspect it.

Contents:

Supplementary document is the Supplementary material of our paper.

In the folder, we don't put the data because we are not allowed to share it. We give the codes to create and estimate Geodesics Bending and assess its performance. Only Synthetic experiment can be performed from the beginning to the end since the others require data.

In the folder Experiment_Alzheimer_cohort, there are all ressources linked with the Alzheimer_cohort experiment. It includes parameters of the models trained and others.

In the folder leaspy, you will find the librairies to create and analyse the models. Geodesics Bending modifies the library by adding a model called "linearb" in the folder "models" and adding some codes in "api.py" to create the fonction "fit_b" (to learn the deformation from the data). Principal functionalities are exposed in the Notebook "Synthetic_Experiment".

If you are interested in the cross validation routines:
leaspy/utils/resampling/Calibrate/calibration.py
leaspy/utils/posterior_analysis/ComparaisonPrediction.py

In the folder Notebook, you will find all you need to inspect the different ressources.

In the folder Synthetic, you will find some saved parameters associated to synthetic experiment.

Notebooks:

To inspect synthetic experiment -> Synthetic Experiment (if you don't want to train them, it is also possible to just load the models)

To inspect Alzheimer_cohort experiment with the models presented in the articles ->
Alzheimer_cohort_results

To launch simulations with others models or to verify results -> Alzheimer_cohort_training_forecast
(first the fixed effect are estimated on 4 folds then the predictions are done on the fifth fold)

The Notebook "Parameters_evolution" analyse the variation of the fixed-effect estimation through steps.