Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



NN,backprop,standardization
Posted:
Nov 17, 2012 3:22 PM


1. Is it necessary to standardize the inputtarget pair data to mean zerro and variance 1 in backpropagation?
2. If yes, why? Most of the books on NN dont mention this step at all
3. Do i need to standardize only input data or both input as well as target data?
4. If i standardize input data before training, then during simulation(finding output) of the network, do i need to feed the network standardized inputs or original inputs?
Thanx chirag



