Feed-forward neural network training using sparse representation

作者：

Highlights：

• Structure optimization for feed-forward neural networks is a significant problem.

• A novel training algorithm was proposed using the concept of sparse representation.

• Proposed algorithm conducts training and structure optimization simultaneously.

• Experiments show better performance compared to state-of-the-art methods.

摘要

•Structure optimization for feed-forward neural networks is a significant problem.•A novel training algorithm was proposed using the concept of sparse representation.•Proposed algorithm conducts training and structure optimization simultaneously.•Experiments show better performance compared to state-of-the-art methods.

论文关键词：Feed-forward neural network,Structure optimization,Sparse representation,Dictionary learning,Multiple measurement vector

论文评审过程：Received 25 May 2018, Revised 22 August 2018, Accepted 22 August 2018, Available online 6 September 2018, Version of Record 20 September 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.08.038