Interrogation Functions

We implement some ready to use interrogation methods from our paper. We implement the interrogation method of Chkrebtii et al (2016), Schober et al (2019) and Kramer et al (2021).

We also implement two other interrogation methods corresponding to Schober et al (2019) and Kramer et al (2021) where we instead use the variance of Chkrebtii et al (2016).

rodeo.interrogate.interrogate_chkrebtii(key, ode_fun, ode_weight, t, mean_state_pred, var_state_pred, kalman_type, **params)[source]

Interrogate method of Chkrebtii et al (2016); DOI: 10.1214/16-BA1017.

Same arguments and returns as interrogate_rodeo().

rodeo.interrogate.interrogate_kramer(key, ode_fun, ode_weight, t, mean_state_pred, var_state_pred, **params)[source]

First order interrogate method of Kramer et al (2021); DOI: https://doi.org/10.48550/arXiv.2110.11812. Assumes off block diagonals are zero. Same arguments and returns as interrogate_rodeo().

rodeo.interrogate.interrogate_rodeo(key, ode_fun, ode_weight, t, mean_state_pred, var_state_pred, **params)[source]

Rodeo interrogation method.

Parameters:

  • key (PRNGKey) – Jax PRNG key.

  • ode_fun (Callable) – Higher order ODE Callable \(W X_t = f(X_t, t, \theta)\) taking arguments \(X\) and \(t\).

  • ode_weight (ndarray(n_block, n_bmeas, n_bstate)) – Weight matrix.

  • t (float) – Time point.

  • mean_state_pred (ndarray(n_block, n_bstate)) – Mean estimate for state at time t given observations from times [a…t-1]; denoted by \(\mu_{t|t-1}\).

  • var_state_pred (ndarray(n_block, n_bstate, n_bstate)) – Covariance of estimate for state at time t given observations from times [a…t-1]; denoted by \(\Sigma_{t|t-1}\).

  • params – Optional model parameters.

Returns:

  • wgt_meas (ndarray(n_block, n_bmeas, n_bstate)): Interrogation weight matrix.

  • mean_meas (ndarray(n_block, n_bmeas)): Interrogation offset.

  • var_meas (ndarray(n_block, n_bmeas, n_bmeas)): Interrogation variance.

Return type:

(tuple)

rodeo.interrogate.interrogate_schober(key, ode_fun, ode_weight, t, mean_state_pred, var_state_pred, **params)[source]

Interrogate method of Schober et al (2019); DOI: https://doi.org/10.1007/s11222-017-9798-7.

Same arguments and returns as interrogate_rodeo().