Archiving and Publishing
The University Library's RDM team provides some information about archiving data (opens in a new tab) and publishing data (opens in a new tab).
According to IvI policy (opens in a new tab) (see R3), the code and data associated with any published article should be made available in a long term persistent archive with an associated persisted identifier such as a DOI.
Repositories
Both GitHub (opens in a new tab) and GitLab (opens in a new tab) are common implementations of Git (opens in a new tab) used in the computer science and ML community.
With the UvA/HvA GitLab (opens in a new tab), you have the convenience of:
- logging in with your UvAnetID (collaborators outside of UvA can access shared projects via eduID (opens in a new tab)).
- unlimited private repositories
- "local" servers (somewhere in the Netherlands or even in Amsterdam probably?) and "in-house" support (Research IT (opens in a new tab) team)
(If you are totally new to Git, check out the tutorial Learn Git Branching (opens in a new tab).)
Hugging Face (opens in a new tab) also provides an implementation of Git to host your machine learning models, spaces, and datasets. (You can even obtain a DOI for your Hugging Face archive! See next section.)
DOIs
Why?
What are the benefits of obtaining a DOI (or other persistent identifier) for your archive instead of simply pointing to a GitHub or GitLab repo?
- DOI providers make a commitment to maintaining archives for an extended period of time (as required by our RDM policy).
- DOI URLs can handle redirects if/when the content is moved.
- A DOI points to a clear and specific version of your archive (better for reproducibility) rather than to a repo that may be changing continuously.
How?
Both Figshare and Zenodo can provide a DOI for your archive.
With the UvA/HvA Figshare, you have the additional advantages that:
- your files are stored on ISO certified servers in Germany (in case your data is not allowed to leave the EU),
- you can keep publish under embargo if necessary (the archive is not publicly accessible for a certain period of time or forever),
- you can have an anonymous permanent link (in case of blind review of research submissions).
GitHub has integrations with both Zenodo and Figshare to enable referencing and citing content (opens in a new tab).
Hugging Face now also provides DOIs for models or datasets (opens in a new tab) that are published on Hugging Face Hub.
Below are some tools that could be helpful for archiving and publishing data and code: