Date: Jan 20, 2013 3:14 PM
Author: Abdul Waheed
Subject: newff to Learn from new inputs

Hi all,
I have successfully trained and tested an ANN with newff which learns the 7 colors. When I give a new input to the network, which is not among the 7 colors (VIBGYOR colors), it also tells me which color it most resembles to. Now I want that when I query the network for any new color, not in VIBGYOR colors, and it classifies it correctly, I can give it a name, and it learns that name as well. For example my network gets input FF,00,00 and says it is 6 (Red), its okie, but when I put 00,00,00 i would also like to specify that when it says 1.93 ( almost blue, but not exactly), I correct it and say it is black, and it learns it as well. Next time when I query 00,00,00 it should say black (1.93 is now black). Do I have to train it every time with full training set now inclusive of black, or it can just learn this one, without going again into the complete training program?

Regards,
Spark