【開催中止】講演者の都合により開催が中止となりました。
【Canceled】The event is canceled due to the speaker’s schedule.
第10回 人間知・脳・AIセミナー／CHAIN Academic Seminar #8
「Neural mechanisms of supervised versus reinforcement motor learning」
日時 / Date & Time
2020年2月14日（金）16:00-17:00
February 14, 2020, 16:00-17:00
場所 / Venue
北海道大学 学術交流会館 第３会議室
（札幌市北区北8条西5丁目）
Hokkaido University Conference Hall, Meeting Room3
（Kita-8, Nishi-5, Kita-ku, Sapporo）
講師 / Lecturer
Dr. Sungshin Kim
(Center for Neuroscience Imaging Research, Institute for Basic Science, Sungkyunkwan University）
→講師紹介は人間知・脳・AI研究教育センターウェブサイトに掲載しています。
Details of Lecturer
講演要旨 / Abstract
Most researchers agree on that motor learning can be categorized into two distinct types, motor adaptation and skill learning. They are differentiated whether learners recalibrate well-trained movement to changes in environment (adaptation) or generate novel movement patterns (skill learning). The motor adaptation involves a parametric change of a motor controller driven by sensory-prediction errors potentially computed by internal models. Thus, motor adaptation can be considered as supervised learning based on directional error feedbacks. In contrast, the motor skill learning involves generating novel movement patterns to achieve task goals. For the learning signals, the skill learning requires evaluative feedback such as reward or penalty instead of directional errors, thus involving initial exploration to select motor controllers associated with higher rewards. This type of learning is known as reinforcement learning. In this seminar, I compare and contrast supervised learning and reinforcement learning and present a fMRI study for each type of learning. I discuss their neural substrates, notably cortico-cerebellar network (adaptation) and cortico-striatal network (skill learning). For the former, I introduce simple computational models for motor adaptation with multiple time scales and present a fMRI study based on the model. For the latter, I present a de novo motor skill learning task for fMRI experiment recently developed in my lab, in which participants learned to control a computer cursor by moving their fingers from scratch. Lastly, I briefly introduce other ongoing studies and future research related with motor learning & memory in my lab.
言語 / Language
英語 / English
参加費 / Fee
無料 / Free
申込方法 / Application
事前申込不要。直接会場へお越しください。
Application in advance is not required. Please come directly to the venue.
主催 / Organizer
北海道大学 人間知・脳・AI研究教育センター
the Center for Human Nature, Artificial Intelligence and Neuroscience (CHAIN)
お問い合わせ先 / Contact
北海道大学人間知・脳・AI研究教育センター事務室（文学研究院内205室）
Office of the Center for Human Nature, Artificial Intelligence and Neuroscience (CHAIN)
Tel: 011-706-4049 Email: office@chain.hokudai.ac.jp