Backpropagation is not used to directly create a data base.
However, if you have a data base of inputs and targets, one of the backpropagation functions like fitnet (regression or curvefitting) or patternnet (classification or pattern recognition) is used to NOT ONLY output close approximations to training target vectors when the corresponding training input vectors are presented, BUT, more importantly, generalize to nontraining data.
Backpropagation can be used to create nets for testing whether or not your choice of targets and target coding (real?, max? min? binary? unity sum? etc) is useful.
Hope this helps.
Thank you for formally accepting my answer
Greg
Best Answer