Deep Learning for Brains

Schulz, M.A., Yeo, T., Vogelstein, J., Mourao-Miranada, J., Kather, J., Kording, K., Richards, B.A. and Bzdok, D., 2019. Deep learning for brains?: Different linear and nonlinear scaling in UK Biobank brain images vs. machine-learning datasets. bioRxiv, p.757054.

1. create conda environment
   conda env create --prefix .envs/deeperbrain_public -f env.yaml

2. activate environment
   source activate .envs/deeperbrain_public

3. prepare datasets
   python prepare_datasets.py mnist
   This should work for the publicly available datasets MNIST, Fashion, Tissue (Kather et al. 2019), Superconductivity (Hamidieh et al. 2018). UK Biobank data is not public, but you can find details on our preprocessing in lib/ukbb_preprocessing.

4. run analyses, e.g.:
   python run.py --data mnist --model logisticregression --grid v3

5. aggregate results to csv file
   python aggregate_results.py

6. plot, e.g.:
   python plot.py mnist --grid v3