迁移学习研究和算法综述

Transfer Learning Research and Algorithm Review

迁移学习可以将从数据域学习的知识迁移到目标数据域,有效地解决传统机器学习中训练集和测试集经常不服从独立同分布的问题.对迁移学习的研究价值和经典算法进行了综述,最后对迁移学习的未来研究重点进行探讨.

In recent years, transfer learning has been applied for many areas. Transfer learning is the transfer of knowledge learned from a data domain to a target data domain. It effectively solves the main problems in traditional machine learning： The training set and the test set often do not obey independent and identically distributed. This article summarizes the research value of transfer learning and classical algorithms, and finally points out the future progress of transfer learning.