Mother of all BCI Benchmarks
Build a comprehensive benchmark of popular Brain-Computer Interface (BCI) algorithms applied on an extensive list of freely available EEG datasets.
Disclaimer
This is an open science project that may evolve depending on the need of the community.
Welcome!
First and foremost, Welcome! 🎉 Willkommen! 🎊 Bienvenue! 🎈🎈🎈
Thank you for visiting the Mother of all BCI Benchmark repository.
This document is a hub to give you some information about the project. Jump straight to one of the sections below, or just scroll down to find out more.
- What are we doing? (And why?)
- Installation
- Running
- Supported datasets
- Who are we? n
- Get in touch
- Documentation
- Architecture and main concepts
- Citing MOABB and related publications
What are we doing?
The problem
Brain-Computer Interfaces allow to interact with a computer using brain signals. In this project, we focus mostly on electroencephalographic signals (EEG), that is a very active research domain, with worldwide scientific contributions. Still:
- Reproducible Research in BCI has a long way to go.
- While many BCI datasets are made freely available, researchers do not publish code, and reproducing results required to benchmark new algorithms turns out to be trickier than it should be.
- Performances can be significantly impacted by parameters of the preprocessing steps, toolboxes used and implementation “tricks” that are almost never reported in the literature.
As a result, there is no comprehensive benchmark of BCI algorithms, and newcomers are spending a tremendous amount of time browsing literature to find out what algorithm works best and on which dataset.
The solution
The Mother of all BCI Benchmarks allows to:
- Build a comprehensive benchmark of popular BCI algorithms applied on an extensive list of freely available EEG datasets.
- The code will be made available on github, serving as a reference point for the future algorithmic developments.
- Algorithms can be ranked and promoted on a website, providing a clear picture of the different solutions available in the field.
This project will be successful when we read in an abstract “ … the proposed method obtained a score of 89% on the MOABB (Mother of All BCI Benchmarks), outperforming the state of the art by 5% ...”.
Installation
Pip installation
To use MOABB, you could simply do:
pip install MOABB
See Troubleshooting section if you have a problem.
Manual installation
You could fork or clone the repository and go to the downloaded directory, then run:
- install
poetry(only once per machine):
curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python -
or checkout installation instruction or use conda forge version - (Optional, skip if not sure) Disable automatic environment creation:
poetry config virtualenvs.create false - install all dependencies in one command (have to be run in the project directory):
poetry install
See contributors' guidelines for detailed explanation.
Requirements we use
See pyproject.toml file for full list of dependencies
Running
Verify Installation
To ensure it is running correctly, you can also run
python -m unittest moabb.tests
once it is installed.
Use MOABB
First, you could take a look at our tutorials that cover the most important concepts and use cases. Also, we have a several examples available.
You might be interested in MOABB documentation
Troubleshooting
Currently pip install moabb fails when pip version < 21, e.g. with 20.0.2 due to an idna
package conflict. Newer pip versions resolve this conflict automatically. To fix this you
can upgrade your pip version using: pip install -U pip before installing moabb.
Supported datasets
The list of supported datasets can be found here : https://neurotechx.github.io/moabb/datasets.html
Submit a new dataset
you can submit a new dataset by mentioning it to this issue. The datasets currently on our radar can be seen [here] (https://github.com/NeuroTechX/moabb/wiki/Datasets-Support)
Who are we?
The founders of the Mother of all BCI Benchmarks are Alexander Barachant and Vinay Jayaram. This project is under the umbrella of NeuroTechX, the international community for NeuroTech enthusiasts. The project is currently maintained by Sylvain Chevallier.
What do we need?
You! In whatever way you can help.
We need expertise in programming, user experience, software sustainability, documentation and technical writing and project management.
We'd love your feedback along the way.
Our primary goal is to build a comprehensive benchmark of popular BCI algorithms applied on an extensive list of freely available EEG datasets, and we're excited to support the professional development of any and all of our contributors. If you're looking to learn to code, try out working collaboratively, or translate your skills to the digital domain, we're here to help.
Get involved
If you think you can help in any of the areas listed above (and we bet you can) or in any of the many areas that we haven't yet thought of (and here we're sure you can) then please check out our contributors' guidelines and our roadmap.
Please note that it's very important to us that we maintain a positive and supportive environment for everyone who wants to participate. When you join us we ask that you follow our code of conduct in all interactions both on and offline.
Contact us
If you want to report a problem or suggest an enhancement, we'd love for you to open an issue at this github repository because then we can get right on it.
For a less formal discussion or exchanging ideas, you can also reach us on the Gitter channel or join our weekly office hours! This an open video meeting happening on a regular basis, please ask the link on the gitter channel. We are also on NeuroTechX slack #moabb channel.
Architecture and Main Concepts
Datasets
A dataset handles and abstracts low-level access to the data. The dataset will read data stored locally, in the format in which they have been downloaded, and will convert them into a MNE raw object. There are options to pool all the different recording sessions per subject or to evaluate them separately.
Paradigm
A paradigm defines how the raw data will be converted to trials ready to be processed by a decoding algorithm. This is a function of the paradigm used, i.e. in motor imagery one can have two-class, multi-class, or continuous paradigms; similarly, different preprocessing is necessary for ERP vs ERD paradigms.
Evaluations
An evaluation defines how we go from trials per subject and session to a generalization statistic (AUC score, f-score, accuracy, etc) -- it can be either within-recording-session accuracy, across-session within-subject accuracy, across-subject accuracy, or other transfer learning settings.
Pipelines
Pipeline defines all steps required by an algorithm to obtain predictions. Pipelines are typically a chain of sklearn compatible transformers and end with a sklearn compatible estimator. See Pipelines for more info.
Statistics and visualization
Once an evaluation has been run, the raw results are returned as a DataFrame. This can be further processed via the following commands to generate some basic visualization and statistical comparisons:
from moabb.analysis import analyze
results = evaluation.process(pipeline_dict)
analyze(results)
Citing MOABB and related publications
To cite MOABB, you could use the following paper:
Vinay Jayaram and Alexandre Barachant. "MOABB: trustworthy algorithm benchmarking for BCIs." Journal of neural engineering 15.6 (2018): 066011. DOI
If you publish a paper using MOABB, please contact us on gitter or open an issue, and we will add your paper to the dedicated wiki page.
Thank You
Thank you so much (Danke schön! Merci beaucoup!) for visiting the project and we do hope that you'll join us on this amazing journey to build a comprehensive benchmark of popular BCI algorithms applied on an extensive list of freely available EEG datasets.