pangeo_fish.hmm.estimator.CachedEstimator

class pangeo_fish.hmm.estimator.CachedEstimator(predictor_factory: callable, sigma: float | None = None, cache: str | PathLike | BaseStore | MutableMapping = None, progress: bool = False)

Estimator to train and predict gaussian random walk hidden markov models

This estimator caches intermediate data to a zarr store, allowing it to compute tracks that wouldn’t fit into memory otherwise, even on very big machines.

Parameters:

• predictor_factory (callable) – Factory for the predictor class. It expects the parameter (“sigma”) as a keyword argument and returns the predictor instance.

• sigma (float, optional) – The primary model parameter: the standard deviation of the distance per time unit traveled by the fish, in the same unit as the grid coordinates.

• cache (str or zarr.Store) – Zarr store to write intermediate results to.

__init__(predictor_factory: callable, sigma: float | None = None, cache: str | PathLike | BaseStore | MutableMapping = None, progress: bool = False) → None

Methods

__init__(predictor_factory[, sigma, cache, ...])
decode(X[, states, mode, spatial_dims, ...])	Decode the state sequence from the selected model and the data
predict_proba(X, *[, cache, spatial_dims, ...])	Predict the state probabilities
score(X, *[, cache, spatial_dims, ...])	Score the fit of the selected model to the data
set_params(**params)	Set the parameters on a new instance
to_dict()

Attributes

cache
progress
sigma
predictor_factory