Fisher information determinant and stochastic complexity for Markov models

Jun'ichi Takeuchi

doi:10.1109/ISIT.2009.5205510

Fisher information determinant and stochastic complexity for Markov models

Jun'ichi Takeuchi

数理情報

研究成果: Chapter in Book/Report/Conference proceeding › Conference contribution

2 被引用数 (Scopus)

抄録

We study Fisher information of stationary Markov models with a finite alphabet. In particular, we derive the Fisher information determinant of expectation parameter η, which is defined as expectation of Markov type. The Fisher information determinant with respect to Markov kernel parameter (conditional probabilities) is easy to find, while it is not so with respect to the expectation parameter η nor the natural parameter θ. Note that θ and η are of special importance for exponential families including Markov models.

本文言語	英語
ホスト出版物のタイトル	2009 IEEE International Symposium on Information Theory, ISIT 2009
ページ	1894-1898
ページ数	5
DOI	https://doi.org/10.1109/ISIT.2009.5205510
出版ステータス	出版済み - 11 19 2009
イベント	2009 IEEE International Symposium on Information Theory, ISIT 2009 - Seoul, 大韓民国継続期間: 6 28 2009 → 7 3 2009

出版物シリーズ

名前	IEEE International Symposium on Information Theory - Proceedings
ISSN（印刷版）	2157-8102

その他

その他	2009 IEEE International Symposium on Information Theory, ISIT 2009
Country	大韓民国
City	Seoul
Period	6/28/09 → 7/3/09

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
Information Systems
Modelling and Simulation
Applied Mathematics

文献へのアクセス

10.1109/ISIT.2009.5205510

フィンガープリント「Fisher information determinant and stochastic complexity for Markov models」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル

Takeuchi, J 2009, Fisher information determinant and stochastic complexity for Markov models. in 2009 IEEE International Symposium on Information Theory, ISIT 2009., 5205510, IEEE International Symposium on Information Theory - Proceedings, pp. 1894-1898, 2009 IEEE International Symposium on Information Theory, ISIT 2009, Seoul, 大韓民国, 6/28/09. https://doi.org/10.1109/ISIT.2009.5205510

@inproceedings{c2ca3f1d5a744b91a138d95f40ac4871,

title = "Fisher information determinant and stochastic complexity for Markov models",

abstract = "We study Fisher information of stationary Markov models with a finite alphabet. In particular, we derive the Fisher information determinant of expectation parameter η, which is defined as expectation of Markov type. The Fisher information determinant with respect to Markov kernel parameter (conditional probabilities) is easy to find, while it is not so with respect to the expectation parameter η nor the natural parameter θ. Note that θ and η are of special importance for exponential families including Markov models.",

author = "Jun'ichi Takeuchi",

year = "2009",

month = nov,

day = "19",

doi = "10.1109/ISIT.2009.5205510",

language = "English",

isbn = "9781424443130",

series = "IEEE International Symposium on Information Theory - Proceedings",

pages = "1894--1898",

booktitle = "2009 IEEE International Symposium on Information Theory, ISIT 2009",

note = "2009 IEEE International Symposium on Information Theory, ISIT 2009 ; Conference date: 28-06-2009 Through 03-07-2009",

}

TY - GEN

T1 - Fisher information determinant and stochastic complexity for Markov models

AU - Takeuchi, Jun'ichi

PY - 2009/11/19

Y1 - 2009/11/19

N2 - We study Fisher information of stationary Markov models with a finite alphabet. In particular, we derive the Fisher information determinant of expectation parameter η, which is defined as expectation of Markov type. The Fisher information determinant with respect to Markov kernel parameter (conditional probabilities) is easy to find, while it is not so with respect to the expectation parameter η nor the natural parameter θ. Note that θ and η are of special importance for exponential families including Markov models.

AB - We study Fisher information of stationary Markov models with a finite alphabet. In particular, we derive the Fisher information determinant of expectation parameter η, which is defined as expectation of Markov type. The Fisher information determinant with respect to Markov kernel parameter (conditional probabilities) is easy to find, while it is not so with respect to the expectation parameter η nor the natural parameter θ. Note that θ and η are of special importance for exponential families including Markov models.

UR - http://www.scopus.com/inward/record.url?scp=70449499588&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70449499588&partnerID=8YFLogxK

U2 - 10.1109/ISIT.2009.5205510

DO - 10.1109/ISIT.2009.5205510

M3 - Conference contribution

AN - SCOPUS:70449499588

SN - 9781424443130

T3 - IEEE International Symposium on Information Theory - Proceedings

SP - 1894

EP - 1898

BT - 2009 IEEE International Symposium on Information Theory, ISIT 2009

T2 - 2009 IEEE International Symposium on Information Theory, ISIT 2009

Y2 - 28 June 2009 through 3 July 2009

ER -