On Stacking Probabilistic Temporal Models with Bidirectional Information Flow

Thomas Geier, Michael Glodek, Georg Layher, Heiko Neumann, Susanne Biundo, Günther Palm
Proceedings of the Eighth International Conference on Probabilistic Graphical Models, PMLR 52:195-206, 2016.

Abstract

We discuss hierarchical combinations of probabilistic models where the upper layer is crafted for predicting time-series data. The combination of models makes the naïve Bayes assumption, stating that the latent variables of the models are independent given the time-indexed label variables. In this setting an additional independence assumption between time steps and mildly inconsistent results are often accepted to make inference computationally feasible. We discuss how the application of approximate inference to the practically intractable joint model instead, shifts the need for these simplifications from model design time to inference time, and the application of loopy belief propagation to the joint model realizes bidirectional communication between models during inference. A first empirical evaluation of the proposed architecture on an activity recognition task demonstrates the benefits of the layered architecture and examines the effects of bidirectional information flow.

Cite this Paper

BibTeX
@InProceedings{pmlr-v52-geier16, title = {On Stacking Probabilistic Temporal Models with Bidirectional Information Flow}, author = {Geier, Thomas and Glodek, Michael and Layher, Georg and Neumann, Heiko and Biundo, Susanne and Palm, Günther}, booktitle = {Proceedings of the Eighth International Conference on Probabilistic Graphical Models}, pages = {195--206}, year = {2016}, editor = {Antonucci, Alessandro and Corani, Giorgio and Campos}, Cassio Polpo}, volume = {52}, series = {Proceedings of Machine Learning Research}, address = {Lugano, Switzerland}, month = {06--09 Sep}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v52/geier16.pdf}, url = {https://proceedings.mlr.press/v52/geier16.html}, abstract = {We discuss hierarchical combinations of probabilistic models where the upper layer is crafted for predicting time-series data. The combination of models makes the naïve Bayes assumption, stating that the latent variables of the models are independent given the time-indexed label variables. In this setting an additional independence assumption between time steps and mildly inconsistent results are often accepted to make inference computationally feasible. We discuss how the application of approximate inference to the practically intractable joint model instead, shifts the need for these simplifications from model design time to inference time, and the application of loopy belief propagation to the joint model realizes bidirectional communication between models during inference. A first empirical evaluation of the proposed architecture on an activity recognition task demonstrates the benefits of the layered architecture and examines the effects of bidirectional information flow.} }
Endnote
%0 Conference Paper %T On Stacking Probabilistic Temporal Models with Bidirectional Information Flow %A Thomas Geier %A Michael Glodek %A Georg Layher %A Heiko Neumann %A Susanne Biundo %A Günther Palm %B Proceedings of the Eighth International Conference on Probabilistic Graphical Models %C Proceedings of Machine Learning Research %D 2016 %E Alessandro Antonucci %E Giorgio Corani %E Cassio Polpo Campos} %F pmlr-v52-geier16 %I PMLR %P 195--206 %U https://proceedings.mlr.press/v52/geier16.html %V 52 %X We discuss hierarchical combinations of probabilistic models where the upper layer is crafted for predicting time-series data. The combination of models makes the naïve Bayes assumption, stating that the latent variables of the models are independent given the time-indexed label variables. In this setting an additional independence assumption between time steps and mildly inconsistent results are often accepted to make inference computationally feasible. We discuss how the application of approximate inference to the practically intractable joint model instead, shifts the need for these simplifications from model design time to inference time, and the application of loopy belief propagation to the joint model realizes bidirectional communication between models during inference. A first empirical evaluation of the proposed architecture on an activity recognition task demonstrates the benefits of the layered architecture and examines the effects of bidirectional information flow.
RIS
TY - CPAPER TI - On Stacking Probabilistic Temporal Models with Bidirectional Information Flow AU - Thomas Geier AU - Michael Glodek AU - Georg Layher AU - Heiko Neumann AU - Susanne Biundo AU - Günther Palm BT - Proceedings of the Eighth International Conference on Probabilistic Graphical Models DA - 2016/08/15 ED - Alessandro Antonucci ED - Giorgio Corani ED - Cassio Polpo Campos} ID - pmlr-v52-geier16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 52 SP - 195 EP - 206 L1 - http://proceedings.mlr.press/v52/geier16.pdf UR - https://proceedings.mlr.press/v52/geier16.html AB - We discuss hierarchical combinations of probabilistic models where the upper layer is crafted for predicting time-series data. The combination of models makes the naïve Bayes assumption, stating that the latent variables of the models are independent given the time-indexed label variables. In this setting an additional independence assumption between time steps and mildly inconsistent results are often accepted to make inference computationally feasible. We discuss how the application of approximate inference to the practically intractable joint model instead, shifts the need for these simplifications from model design time to inference time, and the application of loopy belief propagation to the joint model realizes bidirectional communication between models during inference. A first empirical evaluation of the proposed architecture on an activity recognition task demonstrates the benefits of the layered architecture and examines the effects of bidirectional information flow. ER -
APA
Geier, T., Glodek, M., Layher, G., Neumann, H., Biundo, S. & Palm, G.. (2016). On Stacking Probabilistic Temporal Models with Bidirectional Information Flow. Proceedings of the Eighth International Conference on Probabilistic Graphical Models, in Proceedings of Machine Learning Research 52:195-206 Available from https://proceedings.mlr.press/v52/geier16.html.

Related Material