rts

rts#

Functions#

ParallelIntegratedRTSSmoother(kernel, t_states, ...)

Wrapper for Parallel RTS smoother

make_associative_params(Phi_aug, Q_aug, RESET, ...)

Generate the associative parameters needed for parallel RTS

_combine_per_pair(left, right)

parallel_integrated_rts_smoother(asso_params)

Jax implementation of the parallel RTS smoother algorithm

Module Contents#

smolgp.solvers.integrated.parallel.rts.ParallelIntegratedRTSSmoother(kernel, t_states, stateid, instid, kalman_results)[source]#

Wrapper for Parallel RTS smoother

Parameters:

  • kernel – StateSpaceModel kernel

  • t_states – time coordinates of the states

  • stateid – exposure start/end indicators

  • instid – instrument IDs

  • kalman_results – output from Kalman filter: m_pred, P_pred, m_filter, P_filter

Returns:

g: smoothed means L: smoothed covariances

Return type:

E

smolgp.solvers.integrated.parallel.rts.make_associative_params(Phi_aug, Q_aug, RESET, t_states, stateid, instid, m_pred, P_pred, m_filter, P_filter)[source]#

Generate the associative parameters needed for parallel RTS

See eqns in Section 4B of Sarkka & Garcia-Fernandez (2020)

smolgp.solvers.integrated.parallel.rts._combine_per_pair(left, right)[source]#
smolgp.solvers.integrated.parallel.rts.parallel_integrated_rts_smoother(asso_params)[source]#

Jax implementation of the parallel RTS smoother algorithm for integrated measurements.

See Section 4B of Sarkka & Garcia-Fernandez (2020) for a detailed description of the algorithm and notation, and Section 3.2.4 of Rubenzahl & Hattori et al. (2025) for the integrated case.

Total runtime (span) complexity is O(N/T + logT) where N is the number of time steps and T is the number of parallel threads.