Spectral Learning of Hidden Markov Models from Dynamic and Static Data

Tzu-Kuo Huang, Jeff Schneider
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):630-638, 2013.

Abstract

We develop spectral learning algorithms for Hidden Markov Models that learn not only from time series, or dynamic data but also static data drawn independently from the HMM’s stationary distribution. This is motivated by the fact that static, orderless snapshots are usually easier to obtain than time series in quite a few dynamic modeling tasks. Building on existing spectral learning algorithms, our methods solve convex optimization problems minimizing squared loss on the dynamic data plus a regularization term on the static data. Experiments on synthetic and real human activities data demonstrate better prediction by the proposed method than existing spectral algorithms.

Cite this Paper


BibTeX
@InProceedings{pmlr-v28-huang13, title = {Spectral Learning of Hidden Markov Models from Dynamic and Static Data}, author = {Huang, Tzu-Kuo and Schneider, Jeff}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {630--638}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {3}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/huang13.pdf}, url = {https://proceedings.mlr.press/v28/huang13.html}, abstract = {We develop spectral learning algorithms for Hidden Markov Models that learn not only from time series, or dynamic data but also static data drawn independently from the HMM’s stationary distribution. This is motivated by the fact that static, orderless snapshots are usually easier to obtain than time series in quite a few dynamic modeling tasks. Building on existing spectral learning algorithms, our methods solve convex optimization problems minimizing squared loss on the dynamic data plus a regularization term on the static data. Experiments on synthetic and real human activities data demonstrate better prediction by the proposed method than existing spectral algorithms.} }
Endnote
%0 Conference Paper %T Spectral Learning of Hidden Markov Models from Dynamic and Static Data %A Tzu-Kuo Huang %A Jeff Schneider %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-huang13 %I PMLR %P 630--638 %U https://proceedings.mlr.press/v28/huang13.html %V 28 %N 3 %X We develop spectral learning algorithms for Hidden Markov Models that learn not only from time series, or dynamic data but also static data drawn independently from the HMM’s stationary distribution. This is motivated by the fact that static, orderless snapshots are usually easier to obtain than time series in quite a few dynamic modeling tasks. Building on existing spectral learning algorithms, our methods solve convex optimization problems minimizing squared loss on the dynamic data plus a regularization term on the static data. Experiments on synthetic and real human activities data demonstrate better prediction by the proposed method than existing spectral algorithms.
RIS
TY - CPAPER TI - Spectral Learning of Hidden Markov Models from Dynamic and Static Data AU - Tzu-Kuo Huang AU - Jeff Schneider BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/05/26 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-huang13 PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 3 SP - 630 EP - 638 L1 - http://proceedings.mlr.press/v28/huang13.pdf UR - https://proceedings.mlr.press/v28/huang13.html AB - We develop spectral learning algorithms for Hidden Markov Models that learn not only from time series, or dynamic data but also static data drawn independently from the HMM’s stationary distribution. This is motivated by the fact that static, orderless snapshots are usually easier to obtain than time series in quite a few dynamic modeling tasks. Building on existing spectral learning algorithms, our methods solve convex optimization problems minimizing squared loss on the dynamic data plus a regularization term on the static data. Experiments on synthetic and real human activities data demonstrate better prediction by the proposed method than existing spectral algorithms. ER -
APA
Huang, T. & Schneider, J.. (2013). Spectral Learning of Hidden Markov Models from Dynamic and Static Data. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(3):630-638 Available from https://proceedings.mlr.press/v28/huang13.html.

Related Material