Concept Learning as Search
Initial state = observations of features and judgement of class membership (this is a cup)
Goal state = feature vector(s) that correctly classify unseen instances
Operators add/delete feature to vectors
State space defined by hypotheses of feature vectors