Abstract: The scarcity of labeled data is a critical challenge in industrial process multi-scale modeling, as learning reliable models from limited labeled data and large-scale unlabeled data is ...