Machine Learning Learning by Examples Winston's concept learner Explanation based generalisation Back propagation Discovery - Lenat's AM Rote learning - Samuel

The vast majority of AI research in machine learning has been on learning from examples. Typically this is learning conccept classifications from a sequence of examples. Winston's program which learned a concept of an arch from positive and negative examples was a pioneer in the field. This work was put into a more general framework by Mitchell and Michalski, and finally linked to theorem proving in Explanation Based Generalisation. Here it is possible to justify the generalistion made from a single example. Neural networks, and in particular, back propagation, are a different approach to the same problem of learning classifications. There has also been machine learning work on other types of learning. For example, Lenat's modelling of discovery in mathematics with his AM program. Also, modelling rote learning in Samuel's draughts program.