Learning to Smooth with Bidirectional Predictive State Inference Machines

Wen Sun Carnegie Mellon University, Roberto Capobianco Sapienza University of Rome, Geoffrey J. Gordon Carnegie Mellon University, J. Andrew Bagnell, Byron Boots Georgia Institute of Technology
Proceedings of the 32nd Conference on Uncertainty in Artificial Intelligence, PMLR R14:385-394, 2016.

Abstract

We present the Smoothing Machine (SMACH, pronounced "smash"), a time-series learning algorithm based on chain Conditional Random Fields (CRFs) with latent states. Unlike previous methods, SMACH is designed to optimize prediction performance when we have information from both past and future observations. By leveraging Predictive State Representations (PSRs), we model beliefs about latent states through predictive states-an alternative but equivalent representation that depends directly on observable quantities. Predictive states enable the use of well-developed supervised learning approaches in place of local-optimum-prone methods like EM: we learn regressors or classifiers that can approximate message passing and marginalization in the space of predictive states. We provide theoretical guarantees on smoothing performance and we empirically verify the efficacy of SMACH on two dynamical system benchmarks.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR14-university16j, title = {Learning to Smooth with Bidirectional Predictive State Inference Machines}, author = {University, Wen Sun Carnegie Mellon and Rome, Roberto Capobianco Sapienza University of and University, Geoffrey J. Gordon Carnegie Mellon and Bagnell, J. Andrew and Technology, Byron Boots Georgia Institute of}, booktitle = {Proceedings of the 32nd Conference on Uncertainty in Artificial Intelligence}, pages = {385--394}, year = {2016}, editor = {Ihler, Alexander and Janzing, Dominik}, volume = {R14}, series = {Proceedings of Machine Learning Research}, month = {25--29 Jun}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r14/main/assets/university16j/university16j.pdf}, url = {https://proceedings.mlr.press/r14/university16j.html}, abstract = {We present the Smoothing Machine (SMACH, pronounced "smash"), a time-series learning algorithm based on chain Conditional Random Fields (CRFs) with latent states. Unlike previous methods, SMACH is designed to optimize prediction performance when we have information from both past and future observations. By leveraging Predictive State Representations (PSRs), we model beliefs about latent states through predictive states-an alternative but equivalent representation that depends directly on observable quantities. Predictive states enable the use of well-developed supervised learning approaches in place of local-optimum-prone methods like EM: we learn regressors or classifiers that can approximate message passing and marginalization in the space of predictive states. We provide theoretical guarantees on smoothing performance and we empirically verify the efficacy of SMACH on two dynamical system benchmarks.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T Learning to Smooth with Bidirectional Predictive State Inference Machines %A Wen Sun Carnegie Mellon University %A Roberto Capobianco Sapienza University of Rome %A Geoffrey J. Gordon Carnegie Mellon University %A J. Andrew Bagnell %A Byron Boots Georgia Institute of Technology %B Proceedings of the 32nd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2016 %E Alexander Ihler %E Dominik Janzing %F pmlr-vR14-university16j %I PMLR %P 385--394 %U https://proceedings.mlr.press/r14/university16j.html %V R14 %X We present the Smoothing Machine (SMACH, pronounced "smash"), a time-series learning algorithm based on chain Conditional Random Fields (CRFs) with latent states. Unlike previous methods, SMACH is designed to optimize prediction performance when we have information from both past and future observations. By leveraging Predictive State Representations (PSRs), we model beliefs about latent states through predictive states-an alternative but equivalent representation that depends directly on observable quantities. Predictive states enable the use of well-developed supervised learning approaches in place of local-optimum-prone methods like EM: we learn regressors or classifiers that can approximate message passing and marginalization in the space of predictive states. We provide theoretical guarantees on smoothing performance and we empirically verify the efficacy of SMACH on two dynamical system benchmarks. %Z Reissued by PMLR on 04 October 2026.
APA
University, W.S.C.M., Rome, R.C.S.U.o., University, G.J.G.C.M., Bagnell, J.A. & Technology, B.B.G.I.o.. (2016). Learning to Smooth with Bidirectional Predictive State Inference Machines. Proceedings of the 32nd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R14:385-394 Available from https://proceedings.mlr.press/r14/university16j.html. Reissued by PMLR on 04 October 2026.

Related Material