セミナー案内 Florent Krzakala 氏@10/14

8 views
Skip to first unread message

髙橋昂

unread,
Oct 1, 2025, 8:04:20 AM (6 days ago) Oct 1
to ibi...@googlegroups.com
ibismlのみなさま


東大知の物理学研究センターで助教をしております髙橋昂と申します。

同所属の樺島祥介教授の代理で、以下のセミナーの案内を投稿いたします。
どうぞよろしくお願いいたします。

----------------------------------------------------------------------------------------------------------------
樺島@東大知の物理です.
下記の要領で,Florent Krzakala 氏@EPFLのセミナーを行います.
Krzakala 氏は統計力学の機械学習への応用で現在
最もアクティブに活躍している研究者の一人です.
今回は,なぜ,統計力学が機械学習の研究に有用なのか
非専門家向けの講演をお願いしました.
ご都合のつく方は,是非ご参加下さい.
#オンラインの配信は行いません.

----------------------------------------------------------
講演者:Florent Krzakala 氏(EPFL)
日時:2025年10月14日(火)10:30〜11:30
会場:東京大学本郷キャンパス理学部1号館206教室
題目:From Statistical Physics to Machine Learning
概要:Machine learning has rapidly become one of the most transformative
tools in science and technology. Yet behind its practical successes
lies a remarkable conceptual bridge to physics. Many of the challenges
faced by modern learning algorithms—high-dimensional landscapes,
randomness, over-parametrization, and phase transitions—are familiar
territory to statistical physicists. Indeed, related questions have
long been explored in the study of error-correction, constraint
satisfaction problems, and spin glasses. In this talk, I will show how
ideas from disordered systems theory, random matrices, and phase
transitions shed light on the behavior of learning systems, and why
physicists are uniquely well-placed to contribute to the emerging
theory of machine learning.
----------------------------------------------------------------------------------------------------------------

Reply all
Reply to author
Forward
0 new messages