Reference / binference.PopulationEstimator

binference.PopulationEstimator

class binference.PopulationEstimator(prior=None, *, method='auto', selection=None, selection_mode=None, bins=None, n_samples=None, rng=None, engine_kwargs=None)

Population inference with the scikit-learn estimator conventions.

fit(X) infers the population from per-object likelihoods X (a Likelihoods or an (N, K) array) and stores the result in result_. predict_proba(X) gives per-object posteriors over bins for (new) objects under the inferred population, and score(X) the mean log posterior-predictive density per object, which can be used to compare priors or binnings by cross-validation. The estimator follows the scikit-learn get_params/set_params contract without depending on scikit-learn.

Parameters:

priorPopulationPrior, optional

Population prior (default Dirichlet(1.0)).

methodstr

Inference engine (see binference.infer()).

selectionarray_like (K,), callable or Selection, optional

Selection function.

selection_mode{“on_measurement”, “on_truth”}, optional

Required with an array or callable selection.

binsBins, optional

Bins for raw array inputs.

n_samplesint, optional

Draws per chain / in total.

rngint, Generator, or None

Random number generator.

engine_kwargsdict, optional

Extra keyword arguments for the engine.

__init__(prior=None, *, method='auto', selection=None, selection_mode=None, bins=None, n_samples=None, rng=None, engine_kwargs=None)

Methods

__init__([prior, method, selection, ...])
fit(X[, y])	Infer the population from likelihoods X.
get_params([deep])	Constructor parameters (scikit-learn convention).
predict(X)	Most probable bin of each object under the inferred population.
predict_proba(X)	Per-object posterior over bins under the inferred population.
sample([n, rng])	Draws of the population fractions from the fitted posterior.
score(X[, y])	Mean log posterior-predictive density per object (higher is better).
set_params(**params)	Set constructor parameters (scikit-learn convention).

fit(X, y=None)

Infer the population from likelihoods X.

Parameters:

XLikelihoods or array_like (N, K)

Per-object likelihoods.

yNone, optional

Ignored; accepted for scikit-learn compatibility.

get_params(deep=True)

Constructor parameters (scikit-learn convention).

Parameters:

deepbool, optional

Accepted for scikit-learn compatibility; has no effect.

predict(X)

Most probable bin of each object under the inferred population.

Parameters:

XLikelihoods or array_like (N, K)

Likelihoods of the objects.

predict_proba(X)

Per-object posterior over bins under the inferred population.

Parameters:

XLikelihoods or array_like (N, K)

Likelihoods of the objects.

sample(n=1, rng=None)

Draws of the population fractions from the fitted posterior.

Parameters:

nint, optional

Number of population draws returned.

rngint, Generator or None

Random number generator.

score(X, y=None)

Mean log posterior-predictive density per object (higher is better).

Parameters:

XLikelihoods or array_like (N, K)

Per-object likelihoods.

yNone, optional

Ignored; accepted for scikit-learn compatibility.

set_params(**params)

Set constructor parameters (scikit-learn convention).

Parameters:

**params

Constructor parameters and their new values.