Bayesian Learning and Inference in Recurrent Switching Linear Dynamical Systems

@inproceedings{Linderman2017BayesianLA,
  title={Bayesian Learning and Inference in Recurrent Switching Linear Dynamical Systems},
  author={Scott W. Linderman and Matthew J. Johnson and Andrew C. Miller and Ryan P. Adams and David M. Blei and Liam Paninski},
  booktitle={AISTATS},
  year={2017}
}
The main paper introduces a Gibbs sampling algorithm for the recurrent SLDS and its siblings, but it is straightforward to derive a mean field variational inference algorithm as well. From this, we can immediately derive a stochastic variational inference (SVI) [Ho↵man et al., 2013] algorithm for conditionally independent time series. We use a structured mean field approximation on the augmented model, p(z1:T , x1:T ,!1:T , ✓ | y1:T ) ⇡ q(z1:T ) q(x1:T ) q(!1:T ) q(✓; ⌘). The first three… CONTINUE READING

Citations

Publications citing this paper.
SHOWING 1-10 OF 23 CITATIONS

Structure-Exploiting variational inference for recurrent switching linear dynamical systems

  • 2017 IEEE 7th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP)
  • 2017
VIEW 8 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED

Adaptive, locally linear models of complex dynamics

  • Proceedings of the National Academy of Sciences of the United States of America
  • 2019

Related Work Our model can be viewed as a Deep Kalman Filter

Philip Becker-Ehmck, Jan Peters, Patrick van der Smagt
  • 2019

UATING DISCRETE LATENT TEMPORAL STRUCTURE

VIEW 3 EXCERPTS
CITES BACKGROUND & METHODS

Efficient Bayesian Inference of Sigmoidal Gaussian Cox Processes

  • Journal of Machine Learning Research
  • 2018
VIEW 2 EXCERPTS
CITES BACKGROUND & METHODS

Similar Papers

Loading similar papers…