Neural Networks with L-1 Regularizer for Sparse Representation of Input Data
Release Time:2019-03-11 Hits:
Indexed by: Conference Paper
Date of Publication: 2014-01-01
Included Journals: CPCI-S
Page Number: 437-440
Key Words: Neural network; Sparsification; L-1 regularization; Gradient descent method
Abstract: A L-1 regularizer is proposed in this paper for the sparsification of the input data supplied to an artificial neural network. Gradient-descent method is used for solving the resulting optimization problem. Numerical experiments show the efficiency of the algorithm.