Data Structure
We use a very simple 2-dimensional data structure where each row is a different trial-level sample that encompasses both the condition and the EEG time series. When training a gan, the conditions parameter is used to identify which column name corresponds to condition labels. The GAN will also automatically identify the EEG time series as long as they are prefaced with Time and have an order to them: Time1, Time2, ..., TimeN. All other columns of data will be ignored.
| Condition | Time1 | Time2 | Time3 | Time4 | ... |
|---|---|---|---|---|---|
| 0 | 2.28 | -4.44 | -0.98 | -5.67 | ... |
| 1 | 11.67 | 0.66 | 1.43 | 11.62 | ... |
| 1 | 11.90 | 8.67 | 1.85 | 0.73 | ... |
| 0 | 6.73 | 3.63 | 3.80 | 5.63 | ... |
| ... | ... | ... | ... | ... | ... |
Notes:
- The package is currently designed to use one electrode of data per sample; however, our developers version can handle multiple electrodes and this will be included in our next release!
- The package does not consider individual differences of participant and so we do not provide it any participant IDs.
- The time series can be as long as you like with the understanding that computational time is proportional to the length of the time series.
- Data must be a csv files.