### Data Description

Recordings, grouped into 23 cases, were collected from 22 subjects (5 males, ages 3–22; and 17 females, ages 1.5–19). (Case chb21 was obtained 1.5 years after case chb01, from the same female subject.) 

All signals were sampled at 256 samples per second with 16-bit resolution. The dataset contain 23 EEG signals.

In all, these records include 198 seizures (182 in the original set of 23 cases). The seizures were extracted and mixed with some non-seizure samples for this binary classification task.
The dataset consists of ~54K samples of (23, 256)-shaped input representing signals from EEG patients.
Both validation and test are ~8072 samples long


Reference:
Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.