pypomp.functional.dpop¶

pypomp.functional.dpop(struct: PompStruct, thetas_array: Array, J: int, alpha: float, process_weight_index: int, keys: Array) → Array[source]¶

DPOP differentiable particle filter log-likelihood objective.

A pure functional implementation of the DPOP differentiable particle filter, intended for composition within custom JAX loops or higher-order functions.

Warning

This function is experimental. Its API and behavior are subject to change in future releases.

This function is analogous to pypomp.functional.mop() as a fully differentiable objective function for parameter estimation. However, it additionally incorporates a per-interval transition log-weight that is assumed to be stored in one of the state components.

The process log-weight is expected to be accumulated over a single observation interval by the user-specified process model. At the beginning of each interval, the corresponding state component should be reset to zero (this is naturally handled by accumvars).

Parameters:

struct (PompStruct) – Compiled structural representation of the POMP model.
thetas_array (jax.Array) – Array of initial parameters of shape (n_reps, n_params), aligned with the canonical order of struct.param_names.
J (int) – Number of particles.
alpha (float) – Alpha parameter for DPOP.
process_weight_index (int) – Index of the process weight state component.
keys (jax.Array) – Random keys of shape (n_reps, ...).

Returns:

Negative DPOP log-likelihood estimates.

Return type:

jax.Array