Stochastic collapsed variational inference for hidden Markov models

Stochastic variational inference for collapsed models has recently been successfully applied to large scale topic modelling. In this paper, we propose a stochastic collapsed variational inference algorithm for hidden Markov models, in a sequential data setting. Given a collapsed hidden Markov Model,...

Olles dieđut

Bibliográfalaš dieđut
Váldodahkkit: Wang, P, Blunsom, P
Materiálatiipa: Conference item
Almmustuhtton: Neural Information Processing Systems 2015
_version_ 1826286425209831424
author Wang, P
Blunsom, P
author_facet Wang, P
Blunsom, P
author_sort Wang, P
collection OXFORD
description Stochastic variational inference for collapsed models has recently been successfully applied to large scale topic modelling. In this paper, we propose a stochastic collapsed variational inference algorithm for hidden Markov models, in a sequential data setting. Given a collapsed hidden Markov Model, we break its long Markov chain into a set of short subchains. We propose a novel sum-product algorithm to update the posteriors of the subchains, taking into account their boundary transitions due to the sequential dependencies. Our experiments on two discrete datasets show that our collapsed algorithm is scalable to very large datasets, memory efficient and significantly more accurate than the existing uncollapsed algorithm.
first_indexed 2024-03-07T01:43:34Z
format Conference item
id oxford-uuid:97a6319e-04cd-467d-a3c1-96e3c81c0d0c
institution University of Oxford
last_indexed 2024-03-07T01:43:34Z
publishDate 2015
publisher Neural Information Processing Systems
record_format dspace
spelling oxford-uuid:97a6319e-04cd-467d-a3c1-96e3c81c0d0c2022-03-27T00:01:25ZStochastic collapsed variational inference for hidden Markov modelsConference itemhttp://purl.org/coar/resource_type/c_5794uuid:97a6319e-04cd-467d-a3c1-96e3c81c0d0cSymplectic Elements at OxfordNeural Information Processing Systems2015Wang, PBlunsom, PStochastic variational inference for collapsed models has recently been successfully applied to large scale topic modelling. In this paper, we propose a stochastic collapsed variational inference algorithm for hidden Markov models, in a sequential data setting. Given a collapsed hidden Markov Model, we break its long Markov chain into a set of short subchains. We propose a novel sum-product algorithm to update the posteriors of the subchains, taking into account their boundary transitions due to the sequential dependencies. Our experiments on two discrete datasets show that our collapsed algorithm is scalable to very large datasets, memory efficient and significantly more accurate than the existing uncollapsed algorithm.
spellingShingle Wang, P
Blunsom, P
Stochastic collapsed variational inference for hidden Markov models
title Stochastic collapsed variational inference for hidden Markov models
title_full Stochastic collapsed variational inference for hidden Markov models
title_fullStr Stochastic collapsed variational inference for hidden Markov models
title_full_unstemmed Stochastic collapsed variational inference for hidden Markov models
title_short Stochastic collapsed variational inference for hidden Markov models
title_sort stochastic collapsed variational inference for hidden markov models
work_keys_str_mv AT wangp stochasticcollapsedvariationalinferenceforhiddenmarkovmodels
AT blunsomp stochasticcollapsedvariationalinferenceforhiddenmarkovmodels