Building feedforward neural networks with random weights for large scale datasets

作者：

Highlights：

• Give an overall iterative method for NNRW to deal with large scale datasets.

• An efficient ANE-NNRW algorithm is developed.

• ANE-NNRW’s convergence is analyzed and proved theoretically.

• Some efficient comparisons experiments are studied.

摘要

•Give an overall iterative method for NNRW to deal with large scale datasets.•An efficient ANE-NNRW algorithm is developed.•ANE-NNRW’s convergence is analyzed and proved theoretically.•Some efficient comparisons experiments are studied.

论文关键词：Large scale data,Neural networks,Learning,Approximate Newton-type method

论文评审过程：Received 11 August 2017, Revised 27 March 2018, Accepted 5 April 2018, Available online 11 April 2018, Version of Record 25 April 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.04.007