Mixed-norm sparse representation for multi view face recognition

作者：

Highlights：

• We introduce a novel mixed-norm that takes a trade-off between ℓ1-norm and ℓ2,1-norm.

• We use ℓ1-norm norm on the loss function to achieve a robust solution.

• We derive a simple and provably convergent algorithm based on the alternative directions method of multipliers framework.

• Extensive experiments have been done to demonstrate the performance of the proposed method.

摘要

Highlights•We introduce a novel mixed-norm that takes a trade-off between ℓ1-norm and ℓ2,1-norm.•We use ℓ1-norm norm on the loss function to achieve a robust solution.•We derive a simple and provably convergent algorithm based on the alternative directions method of multipliers framework.•Extensive experiments have been done to demonstrate the performance of the proposed method.

论文关键词：Multi-pose face recognition,Sparse representation classification,ADMM,Group sparse representation,Multi-task learning,Joint dynamic sparse representation classification,Unsupervised learning,Convex optimization,Robust face recognition

论文评审过程：Received 6 November 2013, Revised 19 February 2015, Accepted 25 February 2015, Available online 17 March 2015, Version of Record 16 May 2015.

论文官网地址：https://doi.org/10.1016/j.patcog.2015.02.022