The Neural Engineering Data Consortium (NEDC) launched to focus the research community on high-impact, common-interest neural engineering research questions. Critically, the NEDC will also generate and curate massive data sets to support statistically significant data-driven solutions to those problems. Competition-based evaluations on common data will drive progress by incentivizing innovation, especially from unfunded groups that are unable to generate their own data. NEDC's first dataset is the Temple University Hospital EEG Corpus (TUH-EEG), which is the world's largest publicly available database of clinical EEG data. If you can't find something, don't hesitate to contact us.
Despite millions of dollars in research expenditures and decades of work, robust solutions to neural engineering problems remain elusive. We believe that progress can be made by focusing the community on a handful of problems of common interest and importance. To facilitate innovation, groups should have access to common data sets, which should in turn be made large enough to reflect the inherent variability of neural processes. The NEDC is a central community resource with the goals of determining key research problems, generating and curating data, and being an independent arbiter for algorithm testing and evaluation.
A central tenet of the NEDC is the need for massive amounts of common- protocol data, without which it will be difficult or impossible to address certain neural engineering questions with true statistical confidence. With the support of the community, the NEDC independently generates and curates such datasets in greater volumes than what any typical PI would be willing or able to produce.
A Proven Paradigm
The NEDC's approach has been successfully applied in other data-intensive fields such as astronomy, particle physics, and natural language processing. In fact, prize-based crowdsourced signal processing has even been used by corporations to improve predictive technologies. Examples include:
- Netflix corporation: Improved their movie prediction algorithm using prize-based crowdsourcing on a massive dataset that they provided.
- Berlin Brain Computer Interface Contest: Elicited innovative algorithms for neural decoding using prized-based crowdsourcing on limited datasets they provided. Importantly, many of the best performing algorithms came from smaller unfunded groups whose contribution depended on the availability of data.
- Linguistics Data Consortium: Generate, curate, and distribute massive linguistics corpora. Provided unbiased community-wide evaluation of competing language processing algorithms. Widely credited with accelerating language processing research over a twenty year period.