Skip to content

Latest commit

 

History

History
 
 

src

README.md

Install the requiered libraries

pip install -r requirements.txt

Datasets

  • Visit the docs/instructions/ directory to find the pre-processing steps or instructions of each of them, either we processed them or collected from providers.
  • Visit the docs/instructions/PTB/ if you would like to use the PTB-XL ECG class-labels and other metadata for your research.
  • By convenience, all the dataset implemented in our experiment are available in this link.
  • Similarly you can download and store them with the next command. Allow 6.28 GB.
python3 get_data.py

How to use the models

$SSSD^{S4}$ and $SSSD^{SA}$

can be accesed through the command line, use their respective configuration files from the config/ directory. Load and reshape the datasets into the train.py and inference.py files accordly.

$CSDI^{S4}$

can be accesed as a python module in a notebook, with three main attributions:

from CSDIS4 import CSDIS4Imputer
imputer = CSDIS4Imputer()
imputer.train(data, masking, missing_ratio, batch_size) # for training
imputer.load_weights('path_to_model', 'path_to_config') # after training
imputations = imputer.impute(data, mask, number_of_samples) # sampling

Fast experiment - Mujoco dataset 90% random missing

python3 train.py -c config/config_SSSDS4.json
python3 inference.py -c config/config_SSSDS4.json