pypomp.core.results.PanelPompTrainResult¶

class pypomp.core.results.PanelPompTrainResult(method: str, execution_time: float | None, key: Array, timestamp: Timestamp = <factory>, theta: PanelParameters | None = None, shared_traces: DataArray = <factory>, unit_traces: DataArray = <factory>, logLiks: DataArray = <factory>, optimizer: Optimizer = <factory>, J: int = 0, M: int = 0, eta: LearningRate | None = None, alpha: float = 0.97, alpha_cooling: float = 1.0)[source]¶

Bases: PanelPompEstimationTracesMixin, PanelPompBaseResult

Result from PanelPomp.train() method.

Attributes

J: int = 0¶: The number of particles used for filtering.

M: int = 0¶: The number of iterations performed.

alpha: float = 0.97¶: The discount factor for the gradient moving average.

alpha_cooling: float = 1.0¶: The cooling factor for the discount factor.

eta: LearningRate | None = None¶: The learning rate object.

theta: PanelParameters | None = None¶: The panel parameter object used for the computation.

shared_traces: DataArray¶: Shared parameter traces across iterations. Dimensions: (“theta_idx”, “iteration”, “variable”)

unit_traces: DataArray¶: Unit-specific parameter traces across iterations. Dimensions: (“theta_idx”, “iteration”, “unit”, “variable”)

logLiks: DataArray¶: Log-likelihoods for each unit across iterations. Dimensions: (“theta_idx”, “iteration”, “unit”)

optimizer: Optimizer¶: The optimizer used for training.

method: str¶: The name of the method that produced this result (e.g., ‘pfilter’, ‘mif’).

execution_time: float | None¶: Total execution time in seconds.

key: jax.Array¶: The JAX random key used for this execution.

timestamp: pd.Timestamp¶: The date and time when the result was created.

Methods

`CLL`([average])	Return conditional log-likelihoods as a DataFrame.
`ESS`([average])	Return Effective Sample Size as a DataFrame.
`__init__`(method, execution_time, key, ...)
`merge`(*results)	Merge multiple result objects of the same type.
`print_summary`([n])	Print a summary of this result.
`to_dataframe`([ignore_nan])	Convert panel result to DataFrame.
`traces`()	Return panel result formatted as traces (long format).