登入選單

返回Google圖書搜尋

Constructive Induction: Learning Concepts for Learning

Jörg-Uwe KeitzGesellschaft für Mathematik und Datenverarbeitung

出版	Gesellschaft für Mathematik und Datenverarbeitung mbH, 1991
URL	http://books.google.com.hk/books?id=CFMWHAAACAAJ&hl=&source=gbs_api

註釋

Abstract: "Recently, some psychological studies and newer developments in machine learning seem to converge in that both demand for an interrelated concept structure and point out the need for background knowledge. The question then is: where does the background knowledge come from? Moreover, in which way is it organized in order to allow for learning a particular concept? In this paper, we first link the arguments of psychologists to arguments of machine learning research. We then propose a constructive induction method which leads to the formation of new concepts based on their suitability to define the goal concepts. Concepts are represented by their relations to other concepts.