What question did this study set out to answer?

To optimize the design, training, and feature visualization of deep convolutional neural networks for decoding imagined or executed movement tasks directly from raw EEG data.

August 7, 2017Human Brain MappingOpen Access

Deep learning with convolutional neural networks for EEG decoding and visualization

View Full Paper

Ask AI

Bookmark

To optimize the design, training, and feature visualization of deep convolutional neural networks for decoding imagined or executed movement tasks directly from raw EEG data.
Evaluated multiple deep ConvNet architectures trained end-to-end on raw EEG signals using batch normalization, exponential linear units, and cropped training strategies.
Benchmarked ConvNet performance against the widely used filter bank common spatial patterns (FBCSP) algorithm.
Introduced novel visualization techniques to spatially and spectrally map the causal contributions of learned EEG features across distinct frequency bands.
Deep ConvNets achieved a mean decoding accuracy of 84.0%, outperforming the standard FBCSP algorithm baseline of 82.1%.
Feature visualization showed that the networks autonomously learned to utilize spectral power modulations in the alpha, beta, and high-gamma frequency bands without predefined feature extraction.

Authors

RSRobin Tibor SchirrmeisterUniversity of Freiburg JSJost Tobias SpringenbergGoogle (United States)LFLukas D. J. FiedererUniversity of Freiburg

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Computational evaluation demonstrates deep ConvNets decode raw EEG with 84.0% accuracy during motor tasks, highlighting their utility for end-to-end brain mapping without handcrafted features.

Cite This Study

Schirrmeister et al. (2017) studied this question.

synapsesocial.com/papers/69d8cd97f39dfae3cad17dbd https://doi.org/10.1002/hbm.23730

View Full Paper

Ask AI

Bookmark

Implication

Computational evaluation demonstrates deep ConvNets decode raw EEG with 84.0% accuracy during motor tasks, highlighting their utility for end-to-end brain mapping without handcrafted features.

Deep learning with convolutional neural networks for EEG decoding and visualization

Key Points

Discussion

Implication

Cite This Study

Deep learning with convolutional neural networks for EEG decoding and visualization

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider