Numbering Code	G-INF01 63173 LE10	Year/Term	2022 ・ Second semester
Number of Credits	2	Course Type	Lecture
Target Year		Target Student
Language	English	Day/Period	Mon.3
Instructor name	YAMAMOTO AKIHIRO (Graduate School of Informatics Professor) KOBAYASHI YASUAKI (Graduate School of Informatics Assistant Professor)
Outline and Purpose of the Course	Machine learning now makes string impact to our daily life. In this course we treat machine learning from discrete data and present its mathematical foundations based on formal language theory and theory of computation. Machine learning techniques based on neural networks are suited for real valued vector data, but are not always for discrete structured data. In this course we provide learning mechanism without neural networks. First we introduce elements needed in formalizing machine learning, and then we explain learnability of various classes of formal languages in the models of identification in the limit and learning with queries. We also introduce some results presented recently in computational learning theory, including its relationship with first-order logic as well as with ideals of polynomials. Secondly, we introduce frequent itemset mining from fixed length of bit-vectors. We also give some extensions including mining closed itemsets, mining frequent substring. At last, as machine learning using Boolean functions for bit vectors，we introduce PAC learning and some with topics related to it.
Course Goals	By taking this course, students are expected to understand mathematical foundations of machine learning from string data, tree data, and bit-vectors of a fixed length.
Schedule and Contents	1. Introduction: Machine learning from discrete data Learning pattern languages from String Data 2. Correctness of learning 3. Learning regular languages in the limit 4. Learning regular languages with queries 5. Learning unions of pattern languages 6. Elementary formal systems and learning 7. Learning tree pattern languages 8. Polynomial ideals and learning 9. Frequent itemset mining 10. Formal concept analysis 11. Frequent substring mining and frequent subtree mining 12-15 PAC learning and related topics to it
Evaluation Methods and Policy	Evaluation is based on the submitted reports on the assignments based on Article7 of Graduate School of Informatics Academic Grading Regulations, which will be provided twice during the course.
Course Requirements	Students are assumed to have fundamental knowledge on mathematics, in particular, set theory, and also to be familiar to algorithms.
Study outside of Class (preparation and review)	In addition to preparing for class, Every week, students should review the slides and documents for the lecture which will be available on the lecturer's homepage http://www.iip.ist.i.kyoto-u.ac.jp/member/akihiro/lectures/lectures.html and also in KULASIS or PandA
References, etc.	Grammatical Inference: Learning Automata and Grammars, Colin de la Higuera, (Cambridge University Press ), ISBN:0521763169 計算論的学習, 榊原 康文, 横森 貴, 小林 聡 , (培風館), ISBN:4563014966