For practical applications, data pre-processing is often a necessary step in the development of solution. In the simplest case, pre-processing may take the form of a linear rescaling of the input variables so that they have similar values regardless of the units in which each of these is expressed. More complex pre-processing may involve
reduction of the dimensionality of the input data. The fact that such dimensionality reduction can lead to improved performance may appear somewhat paradoxical, since it
cannot increase the information content of the input data,and in most cases will reduce it. The solution to this paradox is related to the curse of dimensionality.