


lyngby_nn_esoftline - Soft linesearch with the entropic cost function
function [CNew, WNew,VNew, succes, DV, DW] = lyngby_nn_esoftline( ...
X, T, VOld, WOld, Reg, DV, DW, COld, maxIter)
Input: X Input
T Target output
VOld Old input weights
WOld Old output weights
Reg Regularization
DV Step for input weights
DW Step for output weights
COld Old value of costfunction
maxIter Maximal number of iterations
Output: CNew New value for cost function
VNew New input weights
WNew New output weights
succes 0 if unsuccesfull otherwise returning the
number of iterations
DV Step taken for input weights
DW Step taken for output weights
This function will make a soft linesearch until a decrease in
cost is found or the maximal number of iterations is reached.