Generalisation is the way in which systems take specific examples and construct more abstract rules that cover multiple cases. In simple rote learning one takes examples and simply recalls the expected output or response for each example. When the same example is given the learner returns a correct result, but is unable to deal with any input that wasn't in the training set. Techniques for generalisation attempt to seek more abstract patterns, trends or commonalities in the training set to enable them to respond to novel inputs. The term also has a specific meaning in the ACT-R cognitive architecture.
Also known as: generalization
Used in glossary entries: ACT-R, cognitive architecture, rote learning