B0253
Title: Self-supervised learning for physiological time series data
Authors: Nils Strodthoff - Oldenburg University (Germany) [presenting]
Abstract: The recent developments are reviewed in self-supervised representation learning in the domain of time series data, mostly focusing on physiological time series data such as electrocardiography or electroencephalography data. The advantages and disadvantages of different approaches are discussed, ranging from methods adopted from computer vision to methods that exploit the sequential nature of the time series mostly originating from the audio domain. Finally, the impact of such self-supervised representations is demonstrated for possible downstream applications.
