When I trained the network using NaN as don't care values, the individual confusions from the 3rd, 4th and 5th output-target values are much lower than when I trained the network with only the 'success' data. (the 'fail' data, the ones with don't care values, were removed.)
Info
Diese Frage ist geschlossen. Öffnen Sie sie erneut, um sie zu bearbeiten oder zu beantworten.
How to make the network ignore certain values in the targets? (there are 6 neurons, sometimes i need to ignore the first 5)
1 Ansicht (letzte 30 Tage)
Ältere Kommentare anzeigen
Sorry but I have to ask this question again :(
I have to train a feedforward net: 6-5-6 I-H-O topology, tansig and logsig activation functions, binary values as targets.
The sixth binary target indicates a success(1) or fail(0) situation.
When the value of the SIXTH target is 1, the first five values matter.
However, when the value of the SIXTH target is 0, the first five values are "DON'T CARE" values, or insignificant.
What I want to do is train the network with these kind of target values: 110111 100011 000001 #####0 #####0 with # meaning any binary, 0 or 1. I want the training to IGNORE these # values so they will not affect the weight adjustments. BUT I will always need to use the 6th value whether it is a 0 or 1.
If I assign 0's to the #'s, the weights will adjust to the 0 values. If I assign 1's, weights will adjust to the 1's. If I leave them blank they will become NaN's, and weird thing happens that I'm pretty sure is not correct. Can I use the ~ ?
What should I do? How should I modify the target values? I can't find any similar cases to this, sorry.
Thanks a lot!
Antworten (0)
Diese Frage ist geschlossen.
Siehe auch
Produkte
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!