领域自适应研究综述

经典机器学习算法假设训练数据和测试数据具有相同的输入特征空间和数据分布，但在很多现实应用中这一假设通常并不成立，导致经典机器学习算法失效。领域自适应是一种新的机器学习策略，其关键技术在于通过学习新的特征表达来对齐源域和目标域的数据分布，使得在有标签源域中训练的模型可以直接迁移到没有标签的目标域上，且不会引起模型性能的明显下降。介绍领域自适应的定义、分类和代表性算法，讨论基于度量学习和基于对抗学习的两类领域自适应算法。在此基础上，分析领域自适应的典型应用和现存挑战，并对其发展趋势及未来研究方向进行展望。

Review of Studies on Domain Adaptation

Classical machine learning algorithms assume that the training and testing instances share the same input feature space and data distribution.In many real-world applications, however, this assumption cannot be satisfied, resulting in the failure of the algorithms.Domain adaptation is a novel learning paradigm which aligns the data distribution of the source domain and the target domain by learning new feature representations, so that the models trained on the labeled source domain can be transferred to the unlabeled target domain without significant loss of model performance.This paper introduces the definition, classification and representative algorithms of domain adaptation, focusing on the metric learning-based algorithms and the adversarial learning-based algorithms.On this basis, the typical applications and existing challenges of domain adaptation are discussed.The paper also proposes the development trends and possible future research directions.