Presentation | 2010-11-30 Recent Advances in Information-theoretic Learning Theory : Tracking Latent Dynamics Kenji YAMANISHI, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | This paper addresses the issue of detecting changes of latent strucures behind data. The theory of probabilsitic models with latent variales has extensively been explored in the area of statistics and machine learning, and most research on it have mainly focused on the static nature of that model. In this paper we rather consider the case where the probabilisitic model with latent variables changes over time and our main concern is how to track the structural changes of the model. Various approaches have been taken to address this issue. They include modern techniques of information-theoretic learning theory such as "tracking best experts", "switching distributions", "dynamic model selection." They form a novel trend of information theory and learning theory and have been applied to data mining issues such as novelty detection, network structural change detection. This paper overviews recent advances in the theory of detecting changes of latent structures and its applications to real problems. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | latent dynamics / MDL principle / dynamic model selection / switching theory / data mining / novelty detection / anomaly detection / information-theoretic learning theory |
Paper # | IT2010-51 |
Date of Issue |
Conference Information | |
Committee | IT |
---|---|
Conference Date | 2010/11/23(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Information Theory (IT) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Recent Advances in Information-theoretic Learning Theory : Tracking Latent Dynamics |
Sub Title (in English) | |
Keyword(1) | latent dynamics |
Keyword(2) | MDL principle |
Keyword(3) | dynamic model selection |
Keyword(4) | switching theory |
Keyword(5) | data mining |
Keyword(6) | novelty detection |
Keyword(7) | anomaly detection |
Keyword(8) | information-theoretic learning theory |
1st Author's Name | Kenji YAMANISHI |
1st Author's Affiliation | The University of Tokyo() |
Date | 2010-11-30 |
Paper # | IT2010-51 |
Volume (vol) | vol.110 |
Number (no) | 320 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |