Covariance Scattering Transforms

This folder contains the code to reproduce the experiments in the paper Covariance Scattering Transforms, AAAI 2026.

Requirements

• Python 3.11.5
• pip install -r requirements.txt

Datasets

The file data_preprocess.ipynb contains the code to preprocess each dataset. Below the preliminary steps to download the datasets.

ABIDE:

Following instructions at http://preprocessed-connectomes-project.org/abide/, download the file Phenotypic_V1_0b_preprocessed1.csv and place it at the path data/datasets/abide/Phenotypic_V1_0b_preprocessed1.csv

PPMI

From the website https://ida.loni.usc.edu/, download the files FS7_APARC_CTH.csv, Age_at_visit.csv and place them at the path data/datasets/ppmi/.

ADNI

From the website https://ida.loni.usc.edu/, download the files ADSP_PHC_T1_FS_DATADIC and ADSP_PHC_T1_FS and place them in the folder data/datasets/adni

Experiments

To replicate experiments in the paper, run the following scripts. The dataset should be specified inside the scripts.

• stability_exp.py to run stability experiments in Figures 3 and 9
• pruning.py to run the pruning experiments in Figures 4 and 12
• labeled_size_exp.py to run experiments with varying labeled training data size in Figures 5, 10 and 11
• baselines.py to run baselines in Table 4

The best hyperparameter configuration for each experiment is automatically loaded from the folder best_params. The notebook plots.ipynb contains functions to read and plot the results of the experiments.