Numbering Code	U-ENG26 36205 LJ72	Year/Term	2022 ・ Second semester
Number of Credits	2	Course Type	Lecture
Target Year		Target Student
Language	Japanese	Day/Period	Thu.3
Instructor name	ISHII SHIN (Graduate School of Informatics Professor) NISHINO KO (Graduate School of Informatics Professor)
Outline and Purpose of the Course	Teaching sessions aim to help students learn the basics and applications of machine learning. Students are taught theoretical foundations and applications of statistical machine learning methods (in particular, supervised learning, unsupervised learning, and reinforcement learning), which are inductive approaches to solving complex problems.
Course Goals	To acquire knowledge about the basics of machine learning and to deepen one's understanding to a practical level by writing reports that involve programming.
Schedule and Contents	・ Introduction to statistical machine learning (1 session): Explanations are given on the basic concepts of supervised and unsupervised learning with regards to machine learning based on statistical probability theory. (Professor: Ko Nishino) ・ Supervised learning (6 sessions): In relation to supervised learning, students are taught about linear regression involving the least squares method (1 session), and are given an explanation about linear discrimination involving support vector machines (1 session). Following this, students are taught to solve non-linear optimization problems with gradient methods using perceptrons as a subject (1 session), as well as multi-layer perceptrons and the backpropagation learning method used for them (1 session), deep networks centered on convolutional networks (1 session), and progression towards time series represented by LSTM (1 session). (Professor: Ko Nishino) ・ Unsupervised learning and statistical inference (4 sessions): In terms of unsupervised learning, students are taught basic ideas based on the statistical inference performed via probabilistic models (1 session), graphical models and the inference of the maximum likelihood (1 session), Bayesian inference (1 session), and the applications of image processing, etc (1 session). (Professor: Shin Ishii) ・ Reinforcement learning and exploration (3 sessions): With regards to reinforcement learning, which is an autonomous form of control learning based on rewards, students are taught about derivation from dynamic programming (1 session), formulation via probability approximation methods (1 session), and deep reinforcement learning, which has been applied more in recent years (1 session). If there is time, students are also taught the bandit problem (exploration problem). (Professor: Shin Ishii) ・ Application of machine learning in artificial intelligence (1 session): Students are informed about the latest situation regarding the application of machine learning in artificial intelligence. (Professors: Ko Nishino, Shin Ishii)
Evaluation Methods and Policy	[Evaluation method] Marks from exercises in teaching sessions and reports involving programming (80%); evaluation of performance in teaching sessions (20%) Performance in teaching sessions is evaluated based on participation and remarks made in teaching sessions. [Evaluation policy] Achievement targets are evaluated according to the grade evaluation policy of the Faculty of Engineering.
Course Requirements	Students are required to have knowledge of computer software (60370).
Study outside of Class (preparation and review)	Students must work on reports and assignments that involve programming.
Textbooks	Textbooks/References	Others; printouts are used.
References, etc.	Others; Bishop, C., (translated by Motoda, H., and others), Pataan ninshiki to kikai gakushuu jouge - beizu riron ni yoru toukei-teki yosoku, Springer Japan (2007)
Courses delivered by Instructors with Practical Work Experience	分類： A course with practical content delivered by instructors with practical work experience