• 作者： 洪宗貝; 曾憲雄
• 作者服務機構： 高雄工學院資訊管理學系; 交通大學資訊科學系
• 中文摘要： 機器學習對於建立專家系統時的知識擷取相當重要，它可直接從所使用的例子中獲取所要的概念。根據處理例子時方式的不同，機器學習常分成兩類：逐漸式學習策略及批次式學習策略。在逐漸式學習策略之中，版本空問（version space）學習演算法是其中最著名的一個。然而，版本空間學習演算法主要是用於學出連接式的概念（conjunctive concepts）。當所給的例子只能用選言式的概念（disjunctive concepts）表示時，則此學習演算法不能成功地運作。因此，在這篇論文中，我們將改良版本空間學習演算法，使其可以成功地學出選言式的概念出來。我們所提的新演算法稱為逐漸式多版本空問（incremental multiple version spaces）學習演算法，可以逐漸學習的方式將所要的選言式概念求出。此演算法的正確性亦在此論文中被證明。因此，此演算法的提出將有助於提昇版本空間學習策略的學習能力及擴充其應用範圍。
• 英文摘要： Learning general concepts from a set of training instances has become increasingly important forartificial intelligence research on constructing knowledge-based systems. Learning strategies, accordingto their ways of processing training instances, can generally be divided into two classes：incrementallearning strategies and batch learning strategies. Among incremental learning strategies,the“versionspace" learning strategy is one of the most well-known This learning strategy is, however, mainly appliedto learning conjunctive concepts. When the concepts to be learned are disjunctive in form,the versionspacelearning strategy will get a null version space, which cannot correctly represent the desired concepts.In this paper, we have modified the original version space strategy in order to learn disjunctive conceptsThe new proposed version-space-based learning strategy, called the“incremental multiple version spaces”learning strategy, can successfully learn disjunctive concepts in an incremental way.The correctness ofthe algorithm has been proven Implementation has also been carried out on a Pc/AT to verify the proposedlearning strategy.
• 中文關鍵字： machine learning, version space; incremental learning; batch learning; multiple version spaces; disjunctive concepts
• 英文關鍵字： --