Learning the relative structure of EEG Signals using visual correlation variables

This paper is accepted in Basic models of brain and body assembly in neurips 2025.

Steady-state learning (SSL) offers a promising way to study electroencephalography (EEG) representations from unencoded data, reducing the need for expensive annotations for clinical applications such as sleep detection and sleep detection. While current SSL EEG methods mainly use masked reconstruction techniques such as masked auencoders (Mae) that capture spatial temporal patterns, predicting the position of thinking in time to learn the dependence of long signals. We present documented relative shifts or pars as if, a Novel Detecting function that predicts temporal shifts between eeg window winds. Unlike the re-based methods that focus on the recovery of the spatial pattern, the pars encourage the infa Through a complete evaluation of the various EEG Decoding tasks, we show that the converters with peripheral subjection disease always differ from the existing strategies of the background and transfer the active learning settings, establishing a new EEG paradigm of EEG representation.

** Work done during an Apple internship
† Stanford University
‡ California Institute of Technology
§University of Amsterdam