Data normalization is an essential preprocessing step, which is used to rescale values to fit in a specific range. It assures better convergence during backpropagation. In general, data normalization boils down to subtracting the mean of each data point and dividing by its standard deviation.
