第61回統計的機械学習セミナーのご案内 (7/4)

[Kenji Fukumizu]

Ibismlの皆様

 

統数研の福水です．以下の通り，第61回統計的機械学習セミナーをハイブリッド形式で開催いたします．

皆様のご参加をお待ちしております．

The talk will be given in English.

 

オンライン参加を希望される場合は，以下のgoogle form に登録しZoom情報をお受け取りください．

（現地参加の場合は登録不要です）

https://forms.gle/RVB9aYh2GSQfTDjf8

 

Date & Time: July 4 (Thursday)  2024,   16:00--17:30

場所：　統計数理研究所・D棟３階セミナー室５（ハイブリッド）

https://www.ism.ac.jp/access/index_j.html

 

Speaker:   Heishiro Kanagawa  (Newcastle University) https://noukoudashisoup.github.io/

Title: Reinforcement Learning for Adaptive MCMC

Abstract:

An informal observation, made by several authors, is that the adaptive design of a Markov transition kernel has the flavour of a reinforcement learning task. Yet, to-date it has remained unclear how to actually exploit modern reinforcement learning technologies for adaptive MCMC. The aim of this work is to set out a general framework, called Reinforcement Learning Metropolis--Hastings, that is theoretically supported and empirically validated. Our principal focus is on learning fast-mixing Metropolis--Hastings transition kernels, which we cast as deterministic policies and optimise via a policy gradient. Control of the learning rate provably ensures conditions for ergodicity are satisfied. The methodology is used to construct a gradient-free sampler that out-performs a popular gradient-free adaptive Metropolis--Hastings algorithm on ≈90% of tasks in the PosteriorDB benchmark.

 

 

主催：統計数理研究所 先端データサイエンス研究系 統計的機械学習研究センター

連絡先：福水健次