Open
Description
Related to issues #5.
The kind of processing that requires values computed using the whole dataset might cause problem, especially with lazy datasets.
For example, how could we define a preprocessing function that can subtract the mean from each example? One way could be achieve by using closure and yielding.