SPP-EEGNET: An Input-Agnostic Self-supervised EEG Representation Model for Inter-Dataset Transfer Learning
Xiaomin Li · Vangelis Metsis
Abstract
Self-supervised contrastive learning learns inherent features from unlabeled data. In this work, we develop an EEG feature extractor model and train it on a contrastive learning task. Such a model can then be transferred to any new EEG dataset without the need for modifying the target dataset's original dimensionality. The proposed method shows great potential to improve the performance of downstream classification tasks.
Video
Chat is not available.
Successful Page Load