Incremental learning using generative-rehearsal strategy for fault detection and classification

作者：

Highlights：

• We propose a generative-rehearsal strategy for class incremental learning.

• We combine a pseudorehearsal strategy with multiple generative models for each class.

• The proposed method overcomes catastrophic forgetting in incremental learning.

• The proposed method enables incremental learning with imbalanced data.

• The proposed method using generative models shows memory efficiency.

摘要

•We propose a generative-rehearsal strategy for class incremental learning.•We combine a pseudorehearsal strategy with multiple generative models for each class.•The proposed method overcomes catastrophic forgetting in incremental learning.•The proposed method enables incremental learning with imbalanced data.•The proposed method using generative models shows memory efficiency.

论文关键词：Incremental learning,Pseudorehearsal strategy,Generative adversarial networks,Class imbalance,Catastrophic forgetting

论文评审过程：Received 27 November 2019, Revised 7 April 2021, Accepted 23 June 2021, Available online 26 June 2021, Version of Record 9 July 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115477