Abstract: We demonstrate how to use generative adversarial networks to improve the small data problem when training brain-computer-interfaces. The new approach is based on finely graded frequency ...