Benchmarking Validation Methods for Unsupervised Domain Adaptation

This paper compares and ranks 11 UDA validation methods. Validators estimate model accuracy, which makes them an essential component of any UDA train-test pipeline. We rank these validators to indicate which of them are most useful for the purpose of selecting optimal models, checkpoints, and hyperparameters. In addition, we propose and compare new effective validators and significantly improved versions of existing validators. To the best of our knowledge, this large-scale benchmark study is the first of its kind in the UDA field.