This is code for our work to build a multi-output model that predicts both Cardiac and Respiratory abnormal events in the PSG-Audio Dataset
We are using a processed version of the dataset with shared google drive links. Please see our load data time-series repository for code and details on the raw data conversion.
A paper outlining some of the challenges with two differently and highly imbalanced classes is under review.
Currently, we are working to integrate and tune sample_weights individually for the cardiac and respiratory events within the model.