Teacher-Student Consistency For Multi-Source Domain Adaptation
This addresses domain adaptation challenges for machine learning applications where models need to generalize across multiple source domains to an unlabeled target domain, representing an incremental improvement with strong specific gains.
The paper tackles the problem of negative transfer and knowledge fading in Multi-Source Domain Adaptation by proposing the MUST method, which uses teacher-student consistency to improve target inference, resulting in outperforming state-of-the-art methods on benchmarks like digits, text sentiment analysis, and visual-object recognition, sometimes by large margins.
In Multi-Source Domain Adaptation (MSDA), models are trained on samples from multiple source domains and used for inference on a different, target, domain. Mainstream domain adaptation approaches learn a joint representation of source and target domains. Unfortunately, a joint representation may emphasize features that are useful for the source domains but hurt inference on target (negative transfer), or remove essential information about the target domain (knowledge fading). We propose Multi-source Student Teacher (MUST), a novel procedure designed to alleviate these issues. The key idea has two steps: First, we train a teacher network on source labels and infer pseudo labels on the target. Then, we train a student network using the pseudo labels and regularized the teacher to fit the student predictions. This regularization helps the teacher predictions on the target data remain consistent between epochs. Evaluations of MUST on three MSDA benchmarks: digits, text sentiment analysis, and visual-object recognition show that MUST outperforms current SoTA, sometimes by a very large margin. We further analyze the solutions and the dynamics of the optimization showing that the learned models follow the target distribution density, implicitly using it as information within the unlabeled target data.