Code for paper: Patient-Specific Seizure Prediction via Adder Network and Supervised Contrastive Learning

• Title: Patient-Specific Seizure Prediction via Adder Network and Supervised Contrastive Learning
• Authors: Yuchang Zhao, Chang Li, Xiang Liu, Ruobing Qian, Rencheng Song, Xun Chen
• Institution: Hefei University of Technology
• Published in: IEEE Transactions on Neural Systems and Rehabilitation Engineering

• Before running the code, please download the CHBMIT dataset, unzip it and place it into the right directory. Divide the data into 4s segments and save them as hickle files. Each .hickle data file contains the EEG signals and consponding labels of a subject. There are 2 arrays in the file: data and labels. The shape of data is (32, 1, 1024, 22). The shape of label is (32,1).
• Using AddNet-SCL.py to train and test the model (leave one out cross-validation), result will be saved in a /results file.
• The CHBMIT dataset can be found here.
• The usage on Kaggle dataset is the same as above. The Kaggle dataset can be found here.

• Pyhton3.7
• pytorch (1.10.1 version)

• If you have any questions, please contact [email protected]

• huawei-noah/AdderNet
• HobbitLong/SupContrast