Codes and experiments for the following paper:
A. Farnoosh, B. Rezaei, E. Sennesh, Z. Khan, J. Dy, A. Satpute, J. Hutchinson, J. van de Meent, and S, Ostadabbas, "Deep Markov Spatio-Temporal Factorization."
Contact: Amirreza Farnoosh,
Numpy, Scipy, Pytorch, Nibabel, Tqdm, Matplotlib, Sklearn, Json, Pandas
Run the following snippet to restore results from pre-trained checkpoints for Autism dataset in ./fMRI_results folder. A few instances from each dataset are included to help the code run without errors.
You may replace {site} with Caltec, Leuven, MaxMun, NYU_00, SBL_00, Stanfo, Yale_0, USM_00, DSU_0, UM_1_0, or set -exp autism for the full dataset. Here, checkpoint files for Caltec, SBL_00, Stanfo are only included due to storage limitations.
python dmfa_fMRI.py -t 75 -exp autism_{site} -dir ./data_autism/ -smod ./ckpt_fMRI/ -dpath ./fMRI_results/ -restore
or run the following snippet for training with batch size of 10 (full dataset needs to be downloaded and preprocessed/formatted beforehand):
python dmfa_fMRI.py -t 75 -exp autism_{site} -dir ./data_autism/ -smod ./ckpt_fMRI/ -dpath ./fMRI_results/ -bs 10
After downloading the full Autism dataset, run the following snippet to preprocess/format data:
python generate_fMRI_patches.py -T 75 -dir ./path_to_data/ -ext /*.gz -spath ./data_autism/
Run the following snippet to restore results from pre-trained checkpoints for Depression dataset in ./fMRI_results folder. A few instances from the dataset are included to help the code run without errors.
You may replace {ID} with 1, 2, 3, 4, 5. ID 4 corresponds to the first experiment on Depression dataset in the paper. IDs 2, 3 correspond to the second experiment on Depression dataset in the paper. For ID 5 you need access to a larger part of dataset (which is not included here).
python dmfa_fMRI.py -exp depression_{ID} -dir ./data_depression/ -smod ./ckpt_fMRI/ -dpath ./fMRI_results/ -restore
or run the following snippet for training with batch size of 10 (full dataset needs to be downloaded and preprocessed/formatted beforehand):
python dmfa_fMRI.py -exp depression_{ID} -dir ./data_depression/ -smod ./ckpt_fMRI/ -dpath ./fMRI_results/ -bs 10
After downloading the full Depression dataset, run the following snippet to preprocess/format data:
python generate_fMRI_patches_depression.py -T 6 -dir ./path_to_data/ -spath ./data_depression/
If you have the full Depression dataset downloaded, for ID 5 run the following snippet to restore results:
python dmfa_fMRI.py -exp depression_5 -dir ./data_depression/ -smod ./ckpt_fMRI/ -dpath ./fMRI_results/ -restore -c 1 -du 1 -predict
Run the following snippets to restore results from pre-trained checkpoints for traffic dataset. Here, data files for Hangzhou, Birmingham, Guangzhou are included due to storage limitations.
python dmfa_traffic.py -t 18 -k 10 -file ./data_traffic/tensor_b.mat -smod ./ckpt_traffic/b/ -dpath ./Birmingham_results/ -restore -ID birmingham -days 7
python dmfa_traffic.py -t 108 -k 10 -file ./data_traffic/tensor_h.mat -smod ./ckpt_traffic/h/ -dpath ./Hangzhou_results/ -restore -ID hangzhou -days 5
dmfa_traffic.py -t 288 -k 30 -file ./data_traffic/tensor_s.txt -smod ./ckpt_traffic/s/ -dpath ./Seattle_results/ -restore -ID seattle -days 5
python dmfa_traffic.py -t 144 -k 30 -file ./data_traffic/tensor_g.mat -smod ./ckpt_traffic/g/ -dpath ./Guangzhou_results/ -restore -ID guangzhou -days 5
or run the following snippets for training, and then predicting (e.g., Hangzhou, Birmingham datasets):
python dmfa_traffic.py -t 108 -k 10 -epoch 500 -bs 25 -file ./data_traffic/tensor_h.mat -smod ./Hangzhou_results/ -dpath ./Hangzhou_results/ -days 5
python dmfa_traffic.py -t 108 -k 10 -epoch 500 -bs 25 -file ./data_traffic/tensor_b.mat -smod ./Hangzhou_results/ -dpath ./Hangzhou_results/ -days 5 -predict
python dmfa_traffic.py -t 18 -k 10 -epoch 5000 -bs 30 -file ./data_traffic/tensor_b.mat -smod ./Birmingham_results/ -dpath ./Birmingham_results/ -days 7
python dmfa_traffic.py -t 18 -k 10 -epoch 5000 -bs 30 -file ./data_traffic/tensor_b.mat -smod ./Birmingham_results/ -dpath ./Birmingham_results/ -days 7 -predict
Run the following snippet for the toy example.
python dmfa_toy.py
Run the following snippet to restore results from the pre-trained checkpoint for the synthetic experiment in ./synthetic_results folder (synthetic fMRI data is not included due to storage limitations).
python dmfa_synthetic.py
- This code is for non-commertial purpose only.
- For further inquiry please contact: Augmented Cognition Lab at Northeastern University: http://www.northeastern.edu/ostadabbas/