URECA AY18/19 Identification of Autism Spectrum Disorder with Deep Learning

1. Run download_abide_preproc.py

   Use: nohup python download_abide_preproc.py -d func_preproc -p cpac -s filt_global -o <Storage_location>

2. Run calculate_corr_ABIDE.py to get npy files

3. Run classify_file.py to create 2 folders, and place npy files into the corresponding folder(austistic or normal)

4. Run create_data.py to obtain X.npy and Y.npy for training

5. Train using feed_forward_net.py and iterate values of parameters to find best values for parameters

6. Compare results with other models, by running SVM.py

7. With the best values of parameters, run save_ffn_model.py to save the model

8. Run create_SSM.py to produce SSM(including matrix), stored in both CSV and hdf5 formats

9. Run create_K.py to produce K, stored in hdf5 format

10. Run create_eigen.py to produce eigenvalues(stored in both CSV and hdf5) and eigenvectors(stored in hdf5)

11. Run new_PSM.py to generate tables for important eigenvectors

12. Test_model.py is used to find scores for deep learning models

13. DAE1000.py is used to train first autoencoder of 1000 units

14. DAE600.py is used to train second autoencoder of 600 units

15. DAE_1000_600_2.py is used to finetune the autoencoder

16. DAE_1000_600_2_freeze.py is used to finetune the autoencoder, but with the autoencoder weights frozen