| Neural Network Toolbox | |
Mean squared error with regularization or performance derivative function
Syntax
dPerf_dE = dmsereg('e',E,X,perf,PP)
dPerf_dX = dmsereg('x',E,X,perf,PP)
Description
dmsereg is the derivative function for msereg.
dmsereg('d',E,X,perf,PP) takes these arguments,
E - Matrix or cell array of error vector(s).
X - Vector of all weight and bias values.
perf - Network performance (ignored).
PP - mse performance parameter.
where PP defines one performance parameters,
PP.ratio - Relative importance of errors vs. weight and bias values.
and returns the derivative dPerf/dE.
dmsereg('x',E,X,perf) returns the derivative dPerf/dX.
mse has only one performance parameter.
Examples
Here we define an error E and X for a network with one 3-element output and six weight and bias values.
E = {[1; -2; 0.5]};
X = [0; 0.2; -2.2; 4.1; 0.1; -0.2];
Here the ratio performance parameter is defined so that squared errors are 5 times as important as squared weight and bias values.
pp.ratio = 5/(5+1);
Here we calculate the network's performance, and derivatives of performance.
perf = msereg(E,X,pp)
dPerf_dE = dmsereg('e',E,X,perf,pp)
dPerf_dX = dmsereg('x',E,X,perf,pp)
See Also
| | dmse | dnetprod | |