I want to compare my algorithm with some state-of-the-art algorithms for inductive transfer learning in edge computing. More specifically my algorithm is able to recognize distribution shifts in datasets and if necessary transfer knowledge from a model trained on another dataset to label the newly acquired unlabeled data.
I have already tried a simulation with synthetic data but it was not enough for my paper to be accepted, so now i have to test my model with real datasets that do not contain images. Any ideas?