How to train if there are multiple sources?

[Kim Heng]

Suppose we have 3 sources domain and 1 target domain. How should I train DANN in this case? Should I combine all the 3 sources into only 1 source or should I label those 3 domain separately during training?

[Erlend Davidson]

There's a paper for that: "Multi-adversarial domain adapation".
https://pdfs.semanticscholar.org/ffda/dffca8bb57e59db10e5ee702b8f6ccab52cf.pdf