Generally speaking, when training a deep learning model, like MLP, what kind of data pre-processing operation has to be conducted when the input is a numerical sequence.
- It depends on the context of your problem. If it is an sequence of images with not great quality then we need to perform some morphological operation on it e.g.grayscale,binary image etc. Also if the digits are distorted,then we could translate it up-to certain extent for model to learn easily.
If you simply transform the data to have zero mean unit variance Gaussian, be careful to not include your test samples when calculating the actual mean and variance, by doing so you are doing some kind of data snooping thus test error is not reliable, doesn't reflect the true error.