1. You didn't specify if this is
a. regression/curve-fitting (FITNET OR NEWFIT)
or
b. classification/pattern-recognition (PATTERNNET or NEWPR)
The current functions FITNET and PATTERNET are special cases of FEEDFORWARDNET whereas the obsolete functions NEWFIT and NEWPR are special cases of the obsolete function NEWFF.
2. In either case the DEFAULT is to automatically normalize inputs and de-normalize outputs with MAPMINMAX.
3. However, there are options to use either
or
b. No normalization/de-normalization.
So, IT DEPENDS WHAT THE PROGRAM AUTHOR USED.
4. Personally, I prefer to
a. Use MAPSTD BEFORE THE NET for detecting outliers so I can
either modify or remove them.
b. Accept the default MAPMINMAX because I am too lazy to remove
it.
BOTTOM LINE: Your input should have the same summary statistics as the author's data. Then the net will automatically do the rest.
Hope this helps
Thank you for formally accepting my answer
Greg
0 Comments
Sign in to comment.