Reduced Rank Approximations of Transition Matrices

Juan Lin
Proceedings of the Ninth International Workshop on Artificial Intelligence and Statistics, PMLR R4:195-202, 2003.

Abstract

We present various latent variable models for the reduced rank approximation of transition matrices. Two main categories of models, termed Latent Markov Analysis(LMA) models, are introduced. We first address the case where the transition matrix is consistent with a reversible random walk. A more general case is subsequently addressed. Iterative EM-type algorithms are presented for all models. LMA is applied to clustering based on pairwise similarities, where similarities between objects are described probabilistically. In the model, relationships between the inferred clusters are again described probabilistically by the reduced rank transition matrix. LMA simultaneously infers the clusters and abstracts the relationships between them, which can be represented in the form of a weighted graph. Finally, a "targeted" LMA model is introduced where a prior specification of the transition between latent cluster states is incorporated. This provides an algorithm which searches for clusters satisfying pre-specified relationships.

Cite this Paper


BibTeX
@InProceedings{pmlr-vR4-lin03a, title = {Reduced Rank Approximations of Transition Matrices}, author = {Lin, Juan}, booktitle = {Proceedings of the Ninth International Workshop on Artificial Intelligence and Statistics}, pages = {195--202}, year = {2003}, editor = {Bishop, Christopher M. and Frey, Brendan J.}, volume = {R4}, series = {Proceedings of Machine Learning Research}, month = {03--06 Jan}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/r4/lin03a/lin03a.pdf}, url = {https://proceedings.mlr.press/r4/lin03a.html}, abstract = {We present various latent variable models for the reduced rank approximation of transition matrices. Two main categories of models, termed Latent Markov Analysis(LMA) models, are introduced. We first address the case where the transition matrix is consistent with a reversible random walk. A more general case is subsequently addressed. Iterative EM-type algorithms are presented for all models. LMA is applied to clustering based on pairwise similarities, where similarities between objects are described probabilistically. In the model, relationships between the inferred clusters are again described probabilistically by the reduced rank transition matrix. LMA simultaneously infers the clusters and abstracts the relationships between them, which can be represented in the form of a weighted graph. Finally, a "targeted" LMA model is introduced where a prior specification of the transition between latent cluster states is incorporated. This provides an algorithm which searches for clusters satisfying pre-specified relationships.}, note = {Reissued by PMLR on 01 April 2021.} }
Endnote
%0 Conference Paper %T Reduced Rank Approximations of Transition Matrices %A Juan Lin %B Proceedings of the Ninth International Workshop on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2003 %E Christopher M. Bishop %E Brendan J. Frey %F pmlr-vR4-lin03a %I PMLR %P 195--202 %U https://proceedings.mlr.press/r4/lin03a.html %V R4 %X We present various latent variable models for the reduced rank approximation of transition matrices. Two main categories of models, termed Latent Markov Analysis(LMA) models, are introduced. We first address the case where the transition matrix is consistent with a reversible random walk. A more general case is subsequently addressed. Iterative EM-type algorithms are presented for all models. LMA is applied to clustering based on pairwise similarities, where similarities between objects are described probabilistically. In the model, relationships between the inferred clusters are again described probabilistically by the reduced rank transition matrix. LMA simultaneously infers the clusters and abstracts the relationships between them, which can be represented in the form of a weighted graph. Finally, a "targeted" LMA model is introduced where a prior specification of the transition between latent cluster states is incorporated. This provides an algorithm which searches for clusters satisfying pre-specified relationships. %Z Reissued by PMLR on 01 April 2021.
APA
Lin, J.. (2003). Reduced Rank Approximations of Transition Matrices. Proceedings of the Ninth International Workshop on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research R4:195-202 Available from https://proceedings.mlr.press/r4/lin03a.html. Reissued by PMLR on 01 April 2021.

Related Material