msnereg

Purpose

Mean squared normalized error with regularization performance function

Syntax

• perf = msnereg(E,Y,NET,FP)
  dPerf_dy = msnereg('dy',E,Y,NET,perf,FP)
  dPerf_dx = msnereg('dx',E,Y,NET,perf,FP)
  info = msnereg(code)
  

Description

msnereg is a network performance function. It measures the network's performance as the weighted sum of two factors: the mean squared normalized error and the mean squared weights and biases.

The normalization insures that networks with multiple outputs will be trained so that accuracy of each output is treated as equally important. Without normalization outputs with larger values (and therefore larger errors) would be treated as more important.

The minimization of weights and biases forces a network to implement as smooth a function as possible, making it more likely to generalize well.

msnereg(E,Y,NET,FP) takes these arguments,

E	Matrix or cell array of error vectors
Y	Matrix or cell array of output vectors (ignored)
NET	Neural network
FP	Function parameters (ignored)

and returns the mean squared normalized error.

msnereg('dy',E,Y,X,perf,FP) returns the derivative of perf with respect to Y.

msnereg('dx',E,Y,X,perf,FP) returns the derivative of perf with respect to X.

msnereg('name') returns the name of this function.

msnereg('pnames') returns the names of the training parameters.

msnereg('pdefaults') returns the default function parameters.

Examples

Here a two-layer feed-forward network is created with a one-element input ranging from -10 to 10, four hidden neurons.

• net = newff([-10 0 10],[0 0.5 1],4);
  net.performFcn = `msnereg';
  

The network is given a batch of inputs P. The error is calculated by subtracting the output Y from target T. Then the mean squared error is calculated.

• p = [-10 -5 0 5 10];
  t = [0 0 1 1 1];
  y = sim(net,p)
  e = t-y
  perf = msnereg(e,y,net)
  

Network Use

To prepare a custom network to be trained with msnereg, set net.performFcn to 'msnereg'. This automatically sets net.performParam to the empty matrix [], because msne has no performance parameters.

In either case, calling train or adapt results in msne's being used to calculate performance.

See Also

msereg, msne

Provide feedback about this page

msne

nctool

© 1984-2009- The MathWorks, Inc.    -   Site Help   -   Patents   -   Trademarks   -   Privacy Policy   -   Preventing Piracy   -   RSS