ECG Classification with Deep Implicit Layers

Data Preprocessing

Data comes from the MIT-BIH database and is prepared by this project: GitHub.

1. Download train data: link
2. Download test data: link
3. Put them in data/

Model Training and Evaluation

There are three models with a similar number of trainable parameters:

1. 1-block ResNet
2. Neural ODE
3. Deep Equilibrium Network

Simply run the Jupyter Notebook ecg.ipynb. Model training will take a while for the Neural ODE and DEQ models.

Requirements

numpy
pandas
scikit-learn
pytorch
torchinfo
torchdiffeq

File Tree

ecg-neural-ode
├── ecg.ipynb - Notebook for training and testing models 
├── models.py - Model functions
├── utils.py - Other helper functions
└───data
        mitdb_360_test.csv - Training data
        mitdb_360_train.csv - Test data