randomising the order

Terms from Artificial Intelligence: humans at the heart of algorithms

For many machine learning algorithms it is often worth randomising the order of training data before presenting it to the algorithm. For example, if all the examples of a particular class are at the beginning of a data set this might make a clustering algorithm initially create sub-clusters of this first class which may get frozen into the final clusters even when more mixed data is encountered; by randomizing the order this effect is minimised.

Used on page 143