Standardization
Another important step in data preparation is standardizing data. In the previous chapter, for the MNIST data, all pixel values were divided by 255 so that the input data was between 0.0 and 1.0. In our case, we applied min-max normalization, which transforms the data linearly using the following function:
xnew = (x-min(x))/(max(x)-min(x))
Since we already know that min(x) = 0 and max(x)=255, this reduces to the following:
xnew = x / 255.0
The other most popular form of standardization scales the feature so that the mean is 0 and the standard deviation from the mean is 1. This is also known as z-scores, and the formula for it is as follows:
xnew = (x - mean(x)) / std.dev(x)
There are three reasons why we need to perform
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access