基于传递推理的迁移学习

Transfer Learning Based on Transitive Inference

迁移学习是一种新的机器学习方法,它的目标是把从源领域学到的知识迁移到目标领域,但是迁移学习的一个主要局限是迁移知识的目标领域需要和源领域要有直接的关联,否则无法在这两个领域中实现知识迁移.介绍了一种基于传递推理的迁移学习方法,它借助发挥桥梁作用的中间区域实现从源领域到目标领域的迁移学习.该方法能在源领域与目标领域之间仅有少量相关性的情况下实现知识迁移,扩大了迁移学习的应用范围.

Transfer learning is a new machine learning method which aims to use knowledge from source domain to help the learning in a target domain. A major limitation of transfer learning is that the source and target domains should be directly related, otherwise knowledge transfer between them will not work. A transfer learning method based on transitive inference is introduced. This method finds an appropriate domain to bridge the given souroe and target domain, and transfers knowledge among all domains effectively. This method can transfer knowledge when the source and target domain share few relations, and expands the scope of applications of transfer learning.