- commitment to open source
- scalability
- transparency
- modularity
- standardization
- iteration
- avoiding reinvention
- sharing of tools and data
standard format for data and metadata
- build on and expand NWB
- collect metadata at start of GBL experiments
- can use custom tools so long as format is the same
- metadata for processed data includes preprocessing
v.2: tools to streamline metadata creation
standardized preprocessing
- spike sorting and cell segmentation / spike detection
- pipeline built out of modular components
- all tools output to the standard format
- streamlined interfaces for ease of use
v.2: web-based version that runs in the cloud
benchmarking and ground truth
- labs collect ground truth data
- use and improve existing benchmarking systems
- prior art: neurofinder / spikefinder
- labs develop and share new algorithms
cloud storage of processed data
- streamlined uploading tools with format validation
- full metadata required upon upload
- raw data stored locally but available upon request
data browsing
- web-based tools ingest data into a database
- support public / private views with a common interface
- APIs will provide access from programming languages
- interface layer for browsing and querying across labs
v.2: relational queries and analytics
analysis reproducibility
- analysis modules published to GitHub
- executable scripts published to GitHub (e.g. Jupyter)
- use Docker + Binder to make immediately reproducible