Recurrent Kalman Networks: Factorized Inference in High-Dimensional Deep Feature Spaces

TL;DR

Recurrent Kalman Networks (RKNs) introduce a scalable, end-to-end trainable deep filtering approach that explicitly models uncertainty in high-dimensional time-series data, outperforming traditional RNNs in uncertainty estimation and prediction accuracy.

Contribution

The paper presents RKNs, a novel deep filtering architecture that simplifies Kalman updates in high-dimensional spaces, enabling direct end-to-end training without approximations.

Findings

RKNs provide more accurate uncertainty estimates than LSTM and GRU.

RKNs achieve slightly better prediction performance.

RKNs outperform recent generative models on image imputation.

Abstract

In order to integrate uncertainty estimates into deep time-series modelling, Kalman Filters (KFs) (Kalman et al., 1960) have been integrated with deep learning models, however, such approaches typically rely on approximate inference techniques such as variational inference which makes learning more complex and often less scalable due to approximation errors. We propose a new deep approach to Kalman filtering which can be learned directly in an end-to-end manner using backpropagation without additional approximations. Our approach uses a high-dimensional factorized latent state representation for which the Kalman updates simplify to scalar operations and thus avoids hard to backpropagate, computationally heavy and potentially unstable matrix inversions. Moreover, we use locally linear dynamic models to efficiently propagate the latent state to the next time step. The resulting network architecture, which we call Recurrent Kalman Network (RKN), can be used for any time-series data, similar to a LSTM (Hochreiter & Schmidhuber, 1997) but uses an explicit representation of uncertainty. As shown by our experiments, the RKN obtains much more accurate uncertainty estimates than an LSTM or Gated Recurrent Units (GRUs) (Cho et al., 2014) while also showing a slightly improved prediction performance and outperforms various recent generative models on an image imputation task.