Deep brain state classification of MEG data

Abstract

Neuroimaging techniques have shown to be useful when studying the brains activity. This paper uses Magnetoencephalography (MEG) data, provided by the Human Connectome Project (HCP), in combination with various deep artificial neural network models to perform brain decoding. More specifically, here we investigate to which extent can we infer the task performed by a subject based on its MEG data. Three models based on compact convolution, combined convolutional and long short-term architecture as well as a model based on multi-view learning that aims at fusing the outputs of the two stream networks are proposed and examined. These models exploit the spatio-temporal MEG data for learning new representations that are used to decode the relevant tasks across subjects. In order to realize the most relevant features of the input signals, two attention mechanisms, i.e. self and global attention, are incorporated in all the models. The experimental results of cross subject multi-class classification on the studied MEG dataset show that the inclusion of attention improves the generalization of the models across subjects.

Symbolic regression for scientific discovery: an application to wind speed forecasting

“…Here we use the same wind speed dataset as used in [18], which is publicly available 1 . The dataset used originates from the National Climatic Data Center (NCDC) and concerns 5 Danish cities and 4 weather features spanning from 2000 to 2010.…”

“…Some other work also investigated explainability techniques aimed at weather forecasting using deep learning. These techniques include occlusion analysis and score maximization [1,15,24]. The present paper aims at extending the work in [13] and applying it to real weather data that includes a high number of input features.…”

Broad-UNet: Multi-scale feature learning for nowcasting tasks

“…Furthermore, while classical machine learning techniques rely on handcrafted features and domain knowledge, deep learning techniques automatize the extraction of those features. Recent advances in deep learning have shown promising results in diverse research areas such as neuroscience, biomedical signal analysis, weather forecasting and dynamical systems, among others [10,11,12,13,14,15,16,17,18,19]. Convolutional Neural Networks (CNNs) are the most popular algorithms used in computer vision [20], achieving the state-of-the-art in various tasks [21,20,22].…”

Deep multi-stations weather forecasting: explainable recurrent convolutional neural networks

“…While traditional NWP methods aim at extracting useful dynamics from a model or to transfer information between models, the purpose of recent data-driven approaches is to simulate an entire system to predict its future state [27]. Machine learning data driven based models have already been successfully applied in various domains such as healthcare, dynamical systems, biomedical signal analysis, neuroscience among others [20,22,21,18,23,1,4,32,17]. The recent advances of * Corresponding author machine learning models has increased the capability to automatically learn the underlying nonlinear complex patterns of weather dynamics [19,30,31].…”