Open slinderman opened 2 years ago
Use jax.grad to implement a generic extended Kalman filter for linear dynamical systems with linear Gaussian dynamics and nonlinear/non-Gaussian emissions. See chapter 18.5 of Probabilistic Machine Learning by Murphy (2012).
jax.grad
To implement the EKF, we will need the unconstrained parameters of a model. See https://github.com/lindermanlab/ssm-jax/issues/32
Use
jax.grad
to implement a generic extended Kalman filter for linear dynamical systems with linear Gaussian dynamics and nonlinear/non-Gaussian emissions. See chapter 18.5 of Probabilistic Machine Learning by Murphy (2012).To implement the EKF, we will need the unconstrained parameters of a model. See https://github.com/lindermanlab/ssm-jax/issues/32