Four knowledgeable referees support acceptance for the contributions, notably a method for detecting feature shift, and I also recommend acceptance. The problem formulation itself is a contribution, and the proposed method is well-motivated and performs well empirically.