The papers of this issue on machine learning: editorial

Machine learning One of the most active areas of AI research is autonomous knowledge acquisition, or mochine learning. The 1987 Machine Learning Workshop (at Irvine, California) was attended by 200 researchers (it becomes a conference in 1988), and the second European Working Session on Learning (at Bled, Yugoslavia) was also well attended. Not only have there been larger and more frequent meetings, but the rate of publication has increased; machine learning now has its own journal (Langley 1986). Also increasing is the interaction with other research areas. For example, theoretical computer science has recently offered a new perspective: Valiant’s (1984) model, that concepts are “probably approximately correct,” has led to several results about tractable learnability (Keams et al. 1987). Whereas theoretical research tends to make machine learning more rigorous, applications and cross-fertilization tend to make it broader. For example, Clancey’s (1985) view of expert systems as classifiers suggests the mechanized acquisition of classification rules, which is a central issue of machine learning. Machine learning encompasses many domains (e.g., see Mitchell et al. 19860). It appears in many forms, whose properties have been studied, e.g., in Michalski (1983), Holte (1985), Mitchell et al. (19866), DeJong and Mooney (1986), Dietterich (1986), Rendell (1986), Gentner (1987), and Stepp (1987).