I try to understand Multilayer Perceptrons for function regression. As a starting reference I am using the matlab neuronal network toolbox. There I mentioned that the tool maps the input and output values to the range of [-1 1]. Why do they do this? I thought that the inputs need to be decorrelated and free of mean, but they don't do this.
3
-
$\begingroup$ Have a look at this question: stackoverflow.com/questions/4674623/… $\endgroup$– Olivier_s_jCommented Sep 11, 2014 at 8:17
-
$\begingroup$ Also: stats.stackexchange.com/questions/31641/… $\endgroup$– Olivier_s_jCommented Sep 11, 2014 at 8:19
-
$\begingroup$ But the normalization regarding variance and mean has to be done before, right? $\endgroup$– Tik0Commented Sep 11, 2014 at 10:21
Add a comment
|