Recurrent Neural Filters: Learning Independent Bayesian Filtering Steps for Time Series Prediction

By Bryan Lim, Stefan Zohren, Stephen J. Roberts. 2019

51 citations recorded by the source corpus.

Despite the recent popularity of deep generative state space models, few comparisons have been made between network architectures and the inference steps of the Bayesian filtering framework – with most models simultaneously approximating both state transition and update steps with a single recurrent neural network (RNN). In this paper, we introduce the Recurrent Neural Filter (RNF), a novel recurrent autoencoder architecture that learns distinct representations for each Bayesian filtering step, captured by a series of encoders and decoders. Testing this on three real-world time series datasets, we demonstrate that the decoupled representations learnt improve the accuracy of one-step-ahead forecasts while providing realistic uncertainty estimates, and also facilitate multistep prediction through the separation of encoder stages.

Subfields: Artificial Intelligence, Management Science and Operations Research, Probabilistic & Bayesian Methods, Signal Processing, Time-Series & Forecasting

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Stefan Zohren
2. Bryan Lim
3. Stephen J. Roberts
4. W. D. Penny
5. Michael A. Osborne

In Management Science and Operations Research

1. Stefan Zohren
2. Bryan Lim
3. Stephen J. Roberts
4. Weiwei Jiang

In Probabilistic & Bayesian Methods

1. Bryan Lim
2. Stefan Zohren
3. Stephen J. Roberts

In Signal Processing

1. Stefan Zohren
2. Bryan Lim
3. Stephen J. Roberts

In Time-Series & Forecasting

1. Stefan Zohren
2. Bryan Lim
3. Stephen J. Roberts

Citations in the corpus

• Time-series forecasting with deep learning: a survey (2020, background): “Relationship to Bayesian filtering As examined in [39], Bayesian filters [45] and RNNs are both similar in their maintenance of a hidden state which is recursively updated over time.”
• Enhancing Time-Series Momentum Strategies Using Deep Neural Networks (2019, methodology): “Future work includes extensions of the framework presented here to incorporate ways to deal better with nonstationarity in the data, such as using the recently introduced recurrent neural filters (Lim, Zohren, and Roberts 2019).”