Next: Multilayer Networks & Nonlinear
Up: Neural Network Learning
Previous: Differences between GD and
- Appear identical, but PTR is for thresholded perceptron and DR
is for a linear unit (or unthresholed perceptron)
- DR can be used to train a thresholded perceptron, by using
as target values to a linear unit,
, and having the
thresholed unit,
. - If the unthresholed perceptron can be trained to fit these values
perfectly then so can the thresholded perceptron
- If the target values cannot be perfectly fit, then the
thresholded perceptron will be correct whenever the linear unit has
the right sign, but this is not guaranteed to happen
Patricia Jean Riddle
Wed Jun 23 13:06:34 NZST 1999