| Popis: |
Reliable and reproducible machine-learning enabled neuroscience research requires large-scale data sharing and analysis. Essential to the analysis of shared datasets are standardized data organization and metadata formatting, a well-documented automated analysis pipeline, and a comprehensive software framework with a compute environment that can adequately support the research. In this chapter, we introduce the combined Brain Imaging Data Structure (BIDS) and Hierarchical Event Descriptors (HED) frameworks and illustrate their example use through the organization and time course annotation of a publicly shared EEG dataset. We show how the open-source software EEGLAB can operate on data formatted using these standards to perform EEG analysis using a variety of techniques including group-based statistical analysis. Finally, we present a way to exploit freely available high-performance computing resources that allows the application of computationally intensive learning methods to ever larger and more diverse data collections. |
