Sequential Monte Carlo Learning for Time Series Structure Discovery

Feras Saad, Brian Patton, Matthew Douglas Hoffman, Rif A. Saurous, Vikash Mansinghka
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:29473-29489, 2023.

Abstract

This paper presents a new approach to automatically discovering accurate models of complex time series data. Working within a Bayesian nonparametric prior over a symbolic space of Gaussian process time series models, we present a novel structure learning algorithm that integrates sequential Monte Carlo (SMC) and involutive MCMC for highly effective posterior inference. Our method can be used both in "online” settings, where new data is incorporated sequentially in time, and in “offline” settings, by using nested subsets of historical data to anneal the posterior. Empirical measurements on real-world time series show that our method can deliver 10x–100x runtime speedups over previous MCMC and greedy-search structure learning algorithms targeting the same model family. We use our method to perform the first large-scale evaluation of Gaussian process time series structure learning on a prominent benchmark of 1,428 econometric datasets. The results show that our method discovers sensible models that deliver more accurate point forecasts and interval forecasts over multiple horizons as compared to widely used statistical and neural baselines that struggle on this challenging data.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-saad23a, title = {Sequential {M}onte {C}arlo Learning for Time Series Structure Discovery}, author = {Saad, Feras and Patton, Brian and Hoffman, Matthew Douglas and A. Saurous, Rif and Mansinghka, Vikash}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {29473--29489}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/saad23a/saad23a.pdf}, url = {https://proceedings.mlr.press/v202/saad23a.html}, abstract = {This paper presents a new approach to automatically discovering accurate models of complex time series data. Working within a Bayesian nonparametric prior over a symbolic space of Gaussian process time series models, we present a novel structure learning algorithm that integrates sequential Monte Carlo (SMC) and involutive MCMC for highly effective posterior inference. Our method can be used both in "online” settings, where new data is incorporated sequentially in time, and in “offline” settings, by using nested subsets of historical data to anneal the posterior. Empirical measurements on real-world time series show that our method can deliver 10x–100x runtime speedups over previous MCMC and greedy-search structure learning algorithms targeting the same model family. We use our method to perform the first large-scale evaluation of Gaussian process time series structure learning on a prominent benchmark of 1,428 econometric datasets. The results show that our method discovers sensible models that deliver more accurate point forecasts and interval forecasts over multiple horizons as compared to widely used statistical and neural baselines that struggle on this challenging data.} }
Endnote
%0 Conference Paper %T Sequential Monte Carlo Learning for Time Series Structure Discovery %A Feras Saad %A Brian Patton %A Matthew Douglas Hoffman %A Rif A. Saurous %A Vikash Mansinghka %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-saad23a %I PMLR %P 29473--29489 %U https://proceedings.mlr.press/v202/saad23a.html %V 202 %X This paper presents a new approach to automatically discovering accurate models of complex time series data. Working within a Bayesian nonparametric prior over a symbolic space of Gaussian process time series models, we present a novel structure learning algorithm that integrates sequential Monte Carlo (SMC) and involutive MCMC for highly effective posterior inference. Our method can be used both in "online” settings, where new data is incorporated sequentially in time, and in “offline” settings, by using nested subsets of historical data to anneal the posterior. Empirical measurements on real-world time series show that our method can deliver 10x–100x runtime speedups over previous MCMC and greedy-search structure learning algorithms targeting the same model family. We use our method to perform the first large-scale evaluation of Gaussian process time series structure learning on a prominent benchmark of 1,428 econometric datasets. The results show that our method discovers sensible models that deliver more accurate point forecasts and interval forecasts over multiple horizons as compared to widely used statistical and neural baselines that struggle on this challenging data.
APA
Saad, F., Patton, B., Hoffman, M.D., A. Saurous, R. & Mansinghka, V.. (2023). Sequential Monte Carlo Learning for Time Series Structure Discovery. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:29473-29489 Available from https://proceedings.mlr.press/v202/saad23a.html.

Related Material