A Review of Deep Transfer Learning and Recent Advancements

By Mohammadreza Iman et al

Published on Dec. 22, 2022

Read the original document by opening this link in a new tab.

Table of Contents

1 Introduction
2 Deep Learning
3 Deep Transfer Learning (DTL)
4 From Transfer Learning to Deep Transfer Learning, Taxonomy
5 Review of Recent Advancements in DTL

Summary

This paper provides a comprehensive review of Deep Transfer Learning (DTL) and recent advancements in the field. It discusses the challenges faced by deep learning models, such as the dependency on extensive labeled data and training costs, and how DTL aims to address these challenges by reusing knowledge from a source data/task. Various DTL approaches, including Finetuning, Freezing CNN layers, and Progressive learning, are explored. The paper also analyzes the recent advancements in DTL over the past five years, highlighting the practical applications in medical imaging, malware classification, facial emotion recognition, and more. Overall, the limitations of current DTL techniques and potential solutions are discussed, emphasizing the importance of continuous research in the field.