Does Matlab has relative square error available in Neural Network toolbox?

I have seen that MSE, SSE, MAE and SAE are possible training functions of a neural network in Matlab. Does it have relative square error available?

Answer 1

[Greg Heath]

The relative (i.e., NORMALIZED) square error is the ratio of the mean-square-error of the model, MSE, to the mean-square-error of the NAIVE CONSTANT-OUTPUT MODEL, MSE00. To minimize the mse of the latter model, the constant output is just the target mean. Correspondingly, MSE00 is just the average target variance.

 MSE00 = mean(var(target',1))

 Since

 MSE = mse(target-output);

 NMSE = MSE/MSE00   % "N"ormalized, or relative, mse.

The coefficient-of-variation or Rsquared, Rsq (See WKIPEDIA)

 Rsq = 1 - NMSE

is the fraction of target variance that is "explained by the model.

I have zillions of posts in both the NEWSGROUP and ANSWERS using the above variables. In some of them I have gone into more detail than I have here.

Hope this helps.

Thank you for formally accepting my answer

Greg

[Greg Heath]

Most of the time I used R2 instead of Rsq.

 SEARCH            NEWSGROUP      ANSWERS

 NEURAL NMSE        51 HITS        100 HITS
 NEURAL R2         144  HITS       113 HITS

HOPE THIS HELPS.

GREG

Answer 2

[Dave Behera]

The only error functions available in the Neural Network Toolbox are MSE, SSE, MAE and SAE. There is no function for calculating the relative square error.