Neural networks, Machine Learning, Deep Learning

dbinokc · November 2017

I am curious if any experimentation is being done with BCI and interfacing to neural networks.
One thought would be to feed EEG signals into an NN while operating a mouse and see if the
NN could be trained to predict where the mouse should move based on EEG signals. I did some searching on
google and the machine learning reddit and surprisingly saw no mention of trying to use the two together.

Based on what little I know about NN, it would seem if there were a usable signal a NN would be able to tease it
out of the EEG signals.

Razikus · March 2018

Hey.

Currently i'm thinking about it, personally i'm programmer, and I'm interested about BCI from some years.

Aaand i'm planning to do big database of EEG tagged data.

For example set of records about "Brain state when programming"

Or

"Brain state when drinking beer"

And so on. Cause main problem for NN is, that we have no public data.

But main problem is that i have no money to buy Cyton board.

Now i'm setupping my own crowdfunding project on polish website "wspieram.to"

If anyone is interested in - please donate at https://www.paypal.me/vvvgame

(If somebody know polish - https://wspieram.to/eeg?pokaz_projekt=1 )

Greetings

wjcroft · March 2018

Many papers listed here,

https://www.google.com/search?q=neural+network+eeg+classification

retiutut · April 2018

Here is a recent example using a Convolutional Neural Network to classify instances of "bullying" or "no bullying":

https://www.nature.com/articles/s41598-017-17562-0.pdf

This research team had access to a 256-ch EEG cap!

wjcroft · April 2018

AJ @pushtheworld and his NeuroTech Meetup team have been working in this area, and have had a series of ongoing meetings.

https://www.meetup.com/NeuroTechNYC/events/249012309/

https://www.meetup.com/NeuroTechNYC/

https://docs.google.com/document/d/1-z37krFA0mPMdFOA0rhtfSlcymeucYQfwlXuKRj9ttg/edit

https://github.com/NeuroTechX/dl-eeg-playground

https://github.com/robintibor/braindecode

AJ might add a few words here about what kind of states have been recognized.

wjcroft · April 2018

Here's a recent post in the Signal Processing group on LinkedIn,

https://www.linkedin.com/groups/959647/959647-6368393048744820736

Some of the links mentioned on that thread are below.

----

http://brainhack-networks.com/

https://npepjournal.biomedcentral.com/articles/10.1186/s40810-016-0017-0

http://iopscience.iop.org/article/10.1088/1741-2560/14/1/016003/meta

https://www.nature.com/articles/s41598-018-21495-7

wjcroft · January 2019

Yannick Roy (NeuroTechX founder) recently published an excellent review paper covering Deep Learning and EEG,

https://arxiv.org/abs/1901.05498

Deep learning-based electroencephalography analysis: a systematic review

Yannick Roy, Hubert Banville, Isabela Albuquerque, Alexandre Gramfort, Tiago H. Falk, Jocelyn Faubert

(Submitted on 16 Jan 2019 (v1), last revised 20 Jan 2019 (this version, v2))

Electroencephalography (EEG) is a complex signal and can require several years of training to be correctly interpreted. Recently, deep learning (DL) has shown great promise in helping make sense of EEG signals due to its capacity to learn good feature representations from raw data. Whether DL truly presents advantages as compared to more traditional EEG processing approaches, however, remains an open question. In this work, we review 156 papers that apply DL to EEG, published between January 2010 and July 2018, and spanning different application domains such as epilepsy, sleep, brain-computer interfacing, and cognitive and affective monitoring. We extract trends and highlight interesting approaches in order to inform future research and formulate recommendations. Various data items were extracted for each study pertaining to 1) the data, 2) the preprocessing methodology, 3) the DL design choices, 4) the results, and 5) the reproducibility of the experiments. Our analysis reveals that the amount of EEG data used across studies varies from less than ten minutes to thousands of hours. As for the model, 40% of the studies used convolutional neural networks (CNNs), while 14% used recurrent neural networks (RNNs), most often with a total of 3 to 10 layers. Moreover, almost one-half of the studies trained their models on raw or preprocessed EEG time series. Finally, the median gain in accuracy of DL approaches over traditional baselines was 5.4% across all relevant studies. More importantly, however, we noticed studies often suffer from poor reproducibility: a majority of papers would be hard or impossible to reproduce given the unavailability of their data and code. To help the field progress, we provide a list of recommendations for future studies and we make our summary table of DL and EEG papers available and invite the community to contribute.

Also accompanying is a Github repo,

https://github.com/hubertjb/dl-eeg-review

Neural networks, Machine Learning, Deep Learning

Comments