Accelerating Brain Research using Explainable Artificial Intelligence

Abstract

In this demo, we present ExBrainable, an open-source application dedicated to modeling, evaluating and visualizing explainable CNN-based models on EEG data for brain/neuroscience research. We have implemented the functions including EEG data loading, model training, evaluation and parameter visualization. The application is also built with a model base including representative convolutional neural network architectures for users to implement without any programming. With its easy-to-use graphic user interface (GUI), it is completely available for investigators of different disciplines with limited resource and limited programming skill. Starting with preprocessed EEG data, users can quickly train the desired model, evaluate the performance, and finally visualize features learned by the model with no pain.