Table 1. The notations utilized in the cost function.

Notation	Meaning
N	Number of data samples
λ	Weight decay parameter
β	The weight of sparsity penalty term
ρ	Sparsity parameter defining the sparsity level
${\widehat{\rho }}_i$	The average activation of hidden neuron j
sl	Number of neurons in layer l
x(i)	Input feature vector
hw,b(x(i))	Output feature vector
$KL(\rho ||{\widehat{\rho }}_i)$	Kullback-Leibler divergence between ρ and ${\widehat{\rho }}_j$
$W_{ji}^l$	The weight on the connection between neuron j in layer l + 1 and neuron i in layer l