Dear friends
I working in training Neural Networks for forecasting. The data sets are divided into three sets (training, testing and validating). I managed to pre-process the training set set and create the NN model. For testing do i need to pre-process the testing set after training independently ?
.
the answer is "no" ; any statistic used to pre-process the validation or test set must be computed from the training set
for instance, if you normalize your input as zero-mean unit-variance, the mean mu and the variance sigma² must be computed from the training set and used to compute x = (X-mu)/sigma² where x is a variable from the test or validation set
.
The answer is 'No" but its not like you don't have to preprocess the testing or validation data set. All of the data have to be processed using same methods and same values of parameters. Whether you can derive the parameters from training set or any algorithm the test and validation data should be processed in same way.
Dear all
Thanks a lot. Now i have a complete sight about the issue thanks again.
- Badges
- Science topic