A Block Coordinate Descent Proximal Method for Simultaneous Filtering and Parameter Estimation

Ramin Raziperchikolaei, Harish Bhat
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:5380-5388, 2019.

Abstract

We propose and analyze a block coordinate descent proximal algorithm (BCD-prox) for simultaneous filtering and parameter estimation of ODE models. As we show on ODE systems with up to d=40 dimensions, as compared to state-of-the-art methods, BCD-prox exhibits increased robustness (to noise, parameter initialization, and hyperparameters), decreased training times, and improved accuracy of both filtered states and estimated parameters. We show how BCD-prox can be used with multistep numerical discretizations, and we establish convergence of BCD-prox under hypotheses that include real systems of interest.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-raziperchikolaei19a, title = {A Block Coordinate Descent Proximal Method for Simultaneous Filtering and Parameter Estimation}, author = {Raziperchikolaei, Ramin and Bhat, Harish}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {5380--5388}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/raziperchikolaei19a/raziperchikolaei19a.pdf}, url = {https://proceedings.mlr.press/v97/raziperchikolaei19a.html}, abstract = {We propose and analyze a block coordinate descent proximal algorithm (BCD-prox) for simultaneous filtering and parameter estimation of ODE models. As we show on ODE systems with up to d=40 dimensions, as compared to state-of-the-art methods, BCD-prox exhibits increased robustness (to noise, parameter initialization, and hyperparameters), decreased training times, and improved accuracy of both filtered states and estimated parameters. We show how BCD-prox can be used with multistep numerical discretizations, and we establish convergence of BCD-prox under hypotheses that include real systems of interest.} }
Endnote
%0 Conference Paper %T A Block Coordinate Descent Proximal Method for Simultaneous Filtering and Parameter Estimation %A Ramin Raziperchikolaei %A Harish Bhat %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-raziperchikolaei19a %I PMLR %P 5380--5388 %U https://proceedings.mlr.press/v97/raziperchikolaei19a.html %V 97 %X We propose and analyze a block coordinate descent proximal algorithm (BCD-prox) for simultaneous filtering and parameter estimation of ODE models. As we show on ODE systems with up to d=40 dimensions, as compared to state-of-the-art methods, BCD-prox exhibits increased robustness (to noise, parameter initialization, and hyperparameters), decreased training times, and improved accuracy of both filtered states and estimated parameters. We show how BCD-prox can be used with multistep numerical discretizations, and we establish convergence of BCD-prox under hypotheses that include real systems of interest.
APA
Raziperchikolaei, R. & Bhat, H.. (2019). A Block Coordinate Descent Proximal Method for Simultaneous Filtering and Parameter Estimation. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:5380-5388 Available from https://proceedings.mlr.press/v97/raziperchikolaei19a.html.

Related Material