Subclass Graph Embedding and a Marginal Fisher Analysis paradigm

作者：

Highlights：

• Graph Embedding is extended in order to integrate subclass information

• The novel Subclass Graph Embedding framework is proposed.

• The kernelized version of the new framework is presented

• Subclass Graph Embedding encapsulates various subspace learning methods.

• A novel Subclass Marginal Fisher Analysis method is proposed.

摘要

Highlights•Graph Embedding is extended in order to integrate subclass information•The novel Subclass Graph Embedding framework is proposed.•The kernelized version of the new framework is presented•Subclass Graph Embedding encapsulates various subspace learning methods.•A novel Subclass Marginal Fisher Analysis method is proposed.

论文关键词：Dimensionality reduction,Subspace learning,Graph Embedding,Subclass structure

论文评审过程：Received 8 May 2014, Revised 14 April 2015, Accepted 30 May 2015, Available online 18 June 2015, Version of Record 19 August 2015.

论文官网地址：https://doi.org/10.1016/j.patcog.2015.05.027