A Comprehensive Survey of Transfer Learning Techniques and Applications Across Domains

Abstract

Transfer learning refers to the process of transferring the learning from one task to another related task. Transfer learning is especially beneficial in situations where we have limited labeled data in our target task. In this review, we describe core transfer learning concepts, such as fine-tuning, domain adaptation, instance-based learning, and transfer reinforcement learning. This study presents a systematic review of recent studies that cover several applications of transfer learning within a variety of fields, including health care, fault diagnosis, smart systems, and robotics. We discuss the role of different models (e.g., CNNs, LSTMs, Transformers, Progressive Neural Networks) within transfer learning, how explainable AI can be used within transfer learning, and how generative models can help with data efficiency and potentially create more interpretable models. We conduct a review of 18 articles found in peer-reviewed journals and suggest that transfer learning provided benefits in terms of accuracy, reduced training times, and help achieve quicker adaptation in real-world scenarios. We also suggest potential future research directions that may allow for scalable and robust transfer learning.