A novel intrinsically explainable model with semantic manifolds established via transformed priors

作者：

Highlights：

• An IIM merging semantic priors and data-driven learning is proposed for representation learning.

• A theoretical analysis of representation error supports the proposed model.

• A practical convolutional operation is introduced for vision representation.

• The proposed model has competitive performance with DNN, but explainable.

摘要

•An IIM merging semantic priors and data-driven learning is proposed for representation learning.•A theoretical analysis of representation error supports the proposed model.•A practical convolutional operation is introduced for vision representation.•The proposed model has competitive performance with DNN, but explainable.

论文关键词：Explainable artificial intelligence (XAI),Representation learning for vision,Intrinsically interpretable model,Manifold learning,Semantic representation

论文评审过程：Received 24 January 2022, Revised 4 July 2022, Accepted 4 July 2022, Available online 8 July 2022, Version of Record 30 July 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109386