Training and test sets
training set of
S+ positive examples from F+, and
optionally a set S- of negative examples from F-
test set
T+ from F+ where T+ ? S+ = ?, and
optionally T- from F- where T- ? S- = ?
In practice, we may not know all members of F+ and F-
- Thus to construct training & test sets, we can randomly divide an initial set of positive examples into a training set S+ and a test set T+ , similarly for S- and T-
- The goal is to accurately describe “new” members of F+ and F- when we come across them