Memory performance

Without pre-processing

This section ...

Epoch only

Fig X. ...

Epoch & Average functions

Fig X. ...

Conclusions

As expected, the differences in memory usage when not preprocessing data are neglectable since most of the memory is claimed by the returned matrix, and iteration through channels only loads a limited amount of data.

With pre-processing

This section ...

The memory usage depends on the (combination of) preprocessing steps. Currently, preprocessing steps are applied on a per channel basis.

• High pass
  • Channel data is only required once
• Early re-referencing
  • if CAR and speed: will keep each channel data (for after the average to subtract is known), so channel data is required once
  • if CAR and mem: will add the channel data to the average(s), discard (to keep a low memory footprint) and reread each channel later (when the average to subtract is known), so channel data is read twice
• line_noise_removal
• Late re-referencing