The brain coding of multidimensional time series

Abstract

[en] Electroencephalography (EEG) is a widely used technique in the study of brain function. A series of electrodes are placed on the scalp measuring an electric signal from a population of neurons over time. In this work, we will focus on the classification of EEG data. The conventional approach to analyse and classify EEG data employed feature extraction methods. However, deep learning techniques have started to be applied to this task. Among the different architectures, Graph Neural Networks (GNNs) have gained especial attention as EGG data contains complex spatiotemporal relations of high dimensionality, that can be interpreted as a graph. Given the potential of GNNs, we will propose a series of models and try to classify and separate different EEGs into three classes of motivation. The data comes from Cos, Deco, and Gilson, Unpublished, a study focusing on the influence of social motivation during a decision making task. Once the models are trained, we will discus their performance and compare them with the results from the aforementioned study.