Double adjacency graphs-based discriminant neighborhood embedding

作者：

Highlights：

• A supervised dimensionality method DAG-DNE is proposed for classification tasks.

• DAG-DNE constructs two neighbor adjacency graphs to balance Links.

• DAG-DNE keeps the local structure of data and finds a good projection for them.

• Experimental results show the feasibility and effectiveness of DAG-DNE.

摘要

Highlights•A supervised dimensionality method DAG-DNE is proposed for classification tasks.•DAG-DNE constructs two neighbor adjacency graphs to balance Links.•DAG-DNE keeps the local structure of data and finds a good projection for them.•Experimental results show the feasibility and effectiveness of DAG-DNE.

论文关键词：Supervised learning,Discriminant neighborhood embedding,Manifold learning,Face recognition,Double adjacency graphs

论文评审过程：Received 8 January 2014, Revised 3 July 2014, Accepted 30 August 2014, Available online 16 September 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.08.025