The original training set is separated into k equal sized groups
Then one of the groups is used as testing data, and the rest of the groups are used separately as training data
The test group is classified against each testing subgroup and with itself
A value of 10 for k is common