Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data
Recurrent neural networks (RNNs) are nonlinear dynamical models commonly used in the machine learning and dynamical systems literature to represent complex dynamical or sequential relationships between variables. Recently, as deep learning models have become more common, RNNs have been used to forec...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2019-02-01
|
| Series: | Entropy |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1099-4300/21/2/184 |
| _version_ | 1811299901829545984 |
|---|---|
| author | Patrick L. McDermott Christopher K. Wikle |
| author_facet | Patrick L. McDermott Christopher K. Wikle |
| author_sort | Patrick L. McDermott |
| collection | DOAJ |
| description | Recurrent neural networks (RNNs) are nonlinear dynamical models commonly used in the machine learning and dynamical systems literature to represent complex dynamical or sequential relationships between variables. Recently, as deep learning models have become more common, RNNs have been used to forecast increasingly complicated systems. Dynamical spatio-temporal processes represent a class of complex systems that can potentially benefit from these types of models. Although the RNN literature is expansive and highly developed, uncertainty quantification is often ignored. Even when considered, the uncertainty is generally quantified without the use of a rigorous framework, such as a fully Bayesian setting. Here we attempt to quantify uncertainty in a more formal framework while maintaining the forecast accuracy that makes these models appealing, by presenting a Bayesian RNN model for nonlinear spatio-temporal forecasting. Additionally, we make simple modifications to the basic RNN to help accommodate the unique nature of nonlinear spatio-temporal data. The proposed model is applied to a Lorenz simulation and two real-world nonlinear spatio-temporal forecasting applications. |
| first_indexed | 2024-04-13T06:43:31Z |
| format | Article |
| id | doaj.art-0e99bcb47e32428b96831e71f3017a4c |
| institution | Directory Open Access Journal |
| issn | 1099-4300 |
| language | English |
| last_indexed | 2024-04-13T06:43:31Z |
| publishDate | 2019-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Entropy |
| spelling | doaj.art-0e99bcb47e32428b96831e71f3017a4c2022-12-22T02:57:40ZengMDPI AGEntropy1099-43002019-02-0121218410.3390/e21020184e21020184Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal DataPatrick L. McDermott0Christopher K. Wikle1Jupiter Intelligence, Boulder, CO 80302, USADepartment of Statistics, University of Missouri, Columbia, MO 65211, USARecurrent neural networks (RNNs) are nonlinear dynamical models commonly used in the machine learning and dynamical systems literature to represent complex dynamical or sequential relationships between variables. Recently, as deep learning models have become more common, RNNs have been used to forecast increasingly complicated systems. Dynamical spatio-temporal processes represent a class of complex systems that can potentially benefit from these types of models. Although the RNN literature is expansive and highly developed, uncertainty quantification is often ignored. Even when considered, the uncertainty is generally quantified without the use of a rigorous framework, such as a fully Bayesian setting. Here we attempt to quantify uncertainty in a more formal framework while maintaining the forecast accuracy that makes these models appealing, by presenting a Bayesian RNN model for nonlinear spatio-temporal forecasting. Additionally, we make simple modifications to the basic RNN to help accommodate the unique nature of nonlinear spatio-temporal data. The proposed model is applied to a Lorenz simulation and two real-world nonlinear spatio-temporal forecasting applications.https://www.mdpi.com/1099-4300/21/2/184recurrent neural networkBayesian machine learningnonlinear dynamical modelslong-lead forecastingspatial-temporal |
| spellingShingle | Patrick L. McDermott Christopher K. Wikle Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data Entropy recurrent neural network Bayesian machine learning nonlinear dynamical models long-lead forecasting spatial-temporal |
| title | Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data |
| title_full | Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data |
| title_fullStr | Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data |
| title_full_unstemmed | Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data |
| title_short | Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data |
| title_sort | bayesian recurrent neural network models for forecasting and quantifying uncertainty in spatial temporal data |
| topic | recurrent neural network Bayesian machine learning nonlinear dynamical models long-lead forecasting spatial-temporal |
| url | https://www.mdpi.com/1099-4300/21/2/184 |
| work_keys_str_mv | AT patricklmcdermott bayesianrecurrentneuralnetworkmodelsforforecastingandquantifyinguncertaintyinspatialtemporaldata AT christopherkwikle bayesianrecurrentneuralnetworkmodelsforforecastingandquantifyinguncertaintyinspatialtemporaldata |