Disclaimer: Dieser Thread wurde aus dem alten Forum importiert. Daher werden eventuell nicht alle Formatierungen richtig angezeigt. Der ursprüngliche Thread beginnt im zweiten Post dieses Threads.
Licenses for a ML model
Hello everybody,
I have some questions regarding licenses applying for a ML model and our coach asked me to share them here.
Our goal is to automatically categorize files based on their content by applying certain tags to them. To achieve this goal we are considering different ML approaches, some of which require a data set.
My questions:
How do licenses work in this case?
What license applies to the ML model created from the data set assuming it is CC licensed and the code creating and using the model is MIT licensed?
What other licenses could be used for the data set (e.g. ODC-BY)?
Do any licensing problems occur when combining data with different licenses (e.g. different CC licenses)?
Cheers,
markus
Like licenses in general, except that licenses made for open source don’t work well for data or models.
„CC licensed“ data and MIT licensed code does not imply anything about the ML model. Look careful, though, to additional information close to the code like a project website with an FAQ to find a license for the model. If nothing is to be found, you can assume that the model is proprietary and your team’s property. We suggest then an open data license like the ODC-BY license that you mention.
Well, what’s a problem here? Yes, there could be issues depending on what you want. I recommend only mixing ODC-BY licensed data.
Thanks for the diligence and intelligent questions!
Dirk