SSM's Gaussian linear dynamical system (LDS) objects expose a function to compute the marginal likelihood of the data, integrating over the continuous latent states. This function can be automatically differentiated with jax.grad. Use Tensorflow Probability's Hamiltonian Monte Carlo (HMC) functionality to perform Bayesian inference over LDS parameters, using the marginal likelihood and a prior on parameter values.
SSM's Gaussian linear dynamical system (LDS) objects expose a function to compute the marginal likelihood of the data, integrating over the continuous latent states. This function can be automatically differentiated with
jax.grad
. Use Tensorflow Probability's Hamiltonian Monte Carlo (HMC) functionality to perform Bayesian inference over LDS parameters, using the marginal likelihood and a prior on parameter values.