Calculates the gradient of loss function value
Transform loss function from {1, -1} to {0, 1}
Transform loss function from {1, -1} to {0, 1}
Our loss function with {0, 1} labels is max(0, 1 - (2y - 1) (f_w(x))) Therefore the gradient is -(2y - 1)*x
scaled labels
Calculates the loss function value
Calculate the gradient numerically, useful for a testing of gradient functions More about gradient checking: http://cs231n.github.io/neural-networks-3/
Calculate the gradient numerically, useful for a testing of gradient functions More about gradient checking: http://cs231n.github.io/neural-networks-3/
input weights vector
input weights vector
delta parameter for a derivative calculation
computed value of the gradient of the loss function
Hinge Loss http://www1.inf.tu-dresden.de/~ds24/lehre/ml_ws_2013/ml_11_hinge.pdf
Created by rzykov on 31/05/17.