What is it about?
Convolutional neural networks are brain-inspired artificial intelligence methods (part of "deep learning") that are very good at processing natural signals like images or sounds. We use them to decode a person's actions from brain signals recorded with electroencephalography (EEG) and reach results at least as good as established brain-signal-decoding methods. We also show how to make the trained networks understandable, e.g., how one can understand what brain signal features the networks use to determine the person's action.
Featured Image
Why is it important?
Our work is the first to show that deep convolutional neural networks can decode a person's action using the raw, only minimally preprocessed, EEG signal at least as good as established EEG decoding methods. With this, we bring the rapidly developing deep learning world to EEG, hopefully allowing for more deep learning improvements to benefit the EEG field in the future. We exhaustively explore different design choices and training strategies for the networks and thereby create a guide for applying the convolutional networks to EEG. Our novel visualizations allow to better understand what the networks are using from the EEG signal.
Perspectives
I loved to write this article with people with very diverse scientific backgrounds, from more neuroscientific backgrounds to more pure machine-learning/deep-learning backgrounds. Since we have people here that are at the forefront of EEG research and other people at the forefront of deep-learning research, it was very stimulating to integrate the knowledge and contributions of all sides. I hope I was able to merge the fascinating insights of all of our different contributors and create an article you enjoy to read and that gives you new ideas! :) Code is available here: https://github.com/robintibor/braindecode/
Robin Tibor Schirrmeister
University Freiburg - Medical Center
Read the Original
This page is a summary of: Deep learning with convolutional neural networks for EEG decoding and visualization, Human Brain Mapping, August 2017, Wiley,
DOI: 10.1002/hbm.23730.
You can read the full text:
Contributors
The following have contributed to this page
