H = decision trees

• complete - can cover any consistent training set
• therefore, no bias
• therefore, find-S is rote learning
• WE NEED A BIAS!