11 - Editorial. Machine Learning and Discovery机器学习与知识发现.pdfVIP

Machine Learning 1: 363-366, 1986 © 1986 Kluwer Academic Publishers , Boston - Manufactured in The Netherland s Editorial: Machine Learning and Discovery Discovery as learning In everyday language, the terms learning and discovery convey rather different meanings. The former suggests a gradual process, while the latter suggests a more rapid mental event, often involving some form of insight . Learning may lead to an unconscious change in knowledge, while one is always aware that a discovery has been made . The result of learning can be declarative or procedural, while the product of discovery is always declarative . Learning often involves a transfer of knowledge from teacher to student ; in contrast, discovery involves acquiring knowledge from the environment without the aid of a tutor . Finally , all human s and most animals learn from experience, but we reserve the term discovery for the accomplishments of a select few . These boundaries are admittedly vague, but they exist nonetheless . Despite the natura l distinction between these concepts, the field of machine learning has always viewed discovery as one of its concerns. Undoubtedly , one reason for this interest is that , like learning , discovery often involves induction — the act of reasoning from specific facts or data to general rules or laws which provide a general characterization of those facts . Another reason (probably related to the first) is that , historically, researchers in machine discovery have also worked on learning problems and have applied related techniques to these tasks. Thus, cont