The L2,1-norm-based unsupervised optimal feature selection with applications to action recognition
作者：
Highlights：
• A feature selection and classification model is proposed.
• Adjacency graph and projection matrix are simultaneously optimized in our model.
• The model can conduct jointly sparse feature extraction and feature selection.
• The high correlation lying in the video frames is reduced in our model.
• An efficiency algorithm with convergence proof is provided to solve the model.
摘要
•A feature selection and classification model is proposed.•Adjacency graph and projection matrix are simultaneously optimized in our model.•The model can conduct jointly sparse feature extraction and feature selection.•The high correlation lying in the video frames is reduced in our model.•An efficiency algorithm with convergence proof is provided to solve the model.
论文关键词：Feature selection,Sparse representation,Dimensionality reduction,Action recognition
论文评审过程：Received 8 February 2015, Revised 13 April 2016, Accepted 8 June 2016, Available online 18 June 2016, Version of Record 23 June 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2016.06.006
