Predictive State Recurrent Neural Networks

By Carlton Downey, Ahmed S. Hefny, Byron Boots, Geoffrey J. Gordon, Boyue Li. 2017

58 citations recorded by the source corpus.

We present a new model, called Predictive State Recurrent Neural Networks (PSRNNs), for filtering and prediction in dynamical systems. PSRNNs draw on insights from both Recurrent Neural Networks (RNNs) and Predictive State Representations (PSRs), and inherit advantages from both types of models. Like many successful RNN architectures, PSRNNs use (potentially deeply composed) bilinear transfer functions to combine information from multiple sources, so that one source can act as a gate for another. These bilinear functions arise naturally from the connection to state updates in Bayes filters like PSRs, in which observations can be viewed as gating belief states. We show that PSRNNs can be learned effectively by combining backpropogation through time (BPTT) with an initialization based on a statistically consistent learning algorithm for PSRs called two-stage regression (2SR). We also show that PSRNNs can be can be factorized using tensor decomposition, reducing model size and suggesting interesting theoretical connections to existing multiplicative architectures such as LSTMs. We applied PSRNNs to 4 datasets, and showed that we outperform several popular alternative approaches to modeling dynamical systems in all cases.

Subfields: Artificial Intelligence, Computational Mathematics, Statistical and Nonlinear Physics

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. Sergey Levine
2. Pieter Abbeel
3. Chelsea Finn
4. Yoshua Bengio
5. Nando de Freitas
6. Ruslan Salakhutdinov
7. Quoc V. Le
8. Misha Denil

In Computational Mathematics

1. Misha Denil
2. Nando de Freitas
3. Krzysztof Marcin Choromanski

In Statistical and Nonlinear Physics

1. Byron Boots
2. Serkan Cabi
3. Bryan Lim
4. Stephen J. Roberts
5. Misha Denil
6. Nando de Freitas
7. Brandon Amos
8. Krzysztof Marcin Choromanski

Citations in the corpus

• Recurrent Neural Filters: Learning Independent Bayesian Filtering Steps for Time Series Prediction (2019, methodology): “Predictive state RNNs (PSRNN) (Choromanski et al., 2018; Downey et al., 2017; Venkatraman et al., 2017) use an alternative formulation of the Bayes filter, utilising a state representation that corresponds to the statistics of a distribution for future observations.”
• Learning Awareness Models (2018, background): “Several authors have implemented intrinsic motivation, or curiosity based objectives in visual space, through predicting interactions with objects (Pinto et al., 2016), or through predicting summary statistics of the future (Venkatraman et al., 2017; Downey et al., 2017).”
• Initialization matters: Orthogonal Predictive State Recurrent Neural Networks (2018, methodology): “Predictive State Recurrent Neural Networks (PSRNNs) (Downey et al., 2017) are a state-of-the-art approach for modeling time-series data which combine the benefits of probabilistic filters and Recurrent Neural Networks into a single model.”