Remap the normalized data [-1,1] to [0,1]

Hi I have a normalized data which is from [-1,1] with mean 0 and standard deviation 1. I want to map it to [0,1] in order to give to the network. Can anyone tell me here how does this remapping work and will we get any benefit of normalization even after changing the range to [0,1]?

Hi Danish!

Note that taking what you’ve written literally this can only be true if half
of your data is -1 and half is 1. You probably mean something a little

mapped_data = (data + 1) / 2

If you are feeding your data to an already-trained network that was
trained with data in the range [0, 1], you will want to normalize your
data to this range.

If you will be training your network from scratch, your training should
accommodate your [-1, 1] data as easily as it does your data that’s
been mapped to [0, 1] (unless you’re doing something perverse).


K. Frank