6 Aug 2018 | Chuanqi Tan1, Fuchun Sun2, Tao Kong1, Wenchang Zhang1, Chao Yang1, and Chunfang Liu2
This survey paper by Chuanqi Tan, Fuchun Sun, Tao Kong, Wenchang Zhang, Chao Yang, and Chunfang Liu from Tsinghua University provides an in-depth review of deep transfer learning, a critical technique for addressing the challenge of insufficient training data in various domains. The authors define deep transfer learning and categorize it into four main categories: instances-based, mapping-based, network-based, and adversarial-based deep transfer learning. Each category is described with specific techniques and examples, highlighting the unique approaches to leveraging knowledge from source domains to improve performance in target domains. The paper also discusses the importance of deep transfer learning in overcoming the limitations of traditional machine learning methods, particularly in scenarios where large-scale, well-annotated datasets are difficult to obtain. The authors conclude by emphasizing the potential of deep transfer learning to solve challenging problems and its growing importance in the field of deep learning.This survey paper by Chuanqi Tan, Fuchun Sun, Tao Kong, Wenchang Zhang, Chao Yang, and Chunfang Liu from Tsinghua University provides an in-depth review of deep transfer learning, a critical technique for addressing the challenge of insufficient training data in various domains. The authors define deep transfer learning and categorize it into four main categories: instances-based, mapping-based, network-based, and adversarial-based deep transfer learning. Each category is described with specific techniques and examples, highlighting the unique approaches to leveraging knowledge from source domains to improve performance in target domains. The paper also discusses the importance of deep transfer learning in overcoming the limitations of traditional machine learning methods, particularly in scenarios where large-scale, well-annotated datasets are difficult to obtain. The authors conclude by emphasizing the potential of deep transfer learning to solve challenging problems and its growing importance in the field of deep learning.