Majorities help minorities: Hierarchical structure guided transfer learning for few-shot fault recognition

作者：

Highlights：

• A hierarchy guided transfer learning framework for few-shot fault recognition.

• A hierarchical structure of classes can be learned by multi-granularity clustering.

• Filtering out irrelevant categories can overcome the problem of negative transfer.

• Fault feature information can be shared between similar fault categories.

摘要

•A hierarchy guided transfer learning framework for few-shot fault recognition.•A hierarchical structure of classes can be learned by multi-granularity clustering.•Filtering out irrelevant categories can overcome the problem of negative transfer.•Fault feature information can be shared between similar fault categories.

论文关键词：Transfer learning,Fault recognition,Few-shot problem,Hierarchical category structure,Complex systems

论文评审过程：Received 22 April 2021, Revised 22 September 2021, Accepted 19 October 2021, Available online 24 October 2021, Version of Record 6 November 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108383