# Backpropagation Algorithm

### From Ufldl

Line 131: | Line 131: | ||

To train our neural network, we can now repeatedly take steps of gradient descent to reduce our cost function <math>\textstyle J(W,b)</math>. | To train our neural network, we can now repeatedly take steps of gradient descent to reduce our cost function <math>\textstyle J(W,b)</math>. | ||

+ | |||

+ | |||

+ | {{Sparse_Autoencoder}} |