ImpactModel

ImpactModel(kernel, rng_key, inference, *[, ...])

Impact modeling interface: fit, sample, predict, and estimate effects.

Attributes

ImpactModel.inference	The underlying NumPyro inference object.
ImpactModel.kernel	A probabilistic model with NumPyro primitives.
ImpactModel.kernel_spec	The cached KernelSpec or None if not yet built.
ImpactModel.param_input	Parameter name in kernel for the input data.
ImpactModel.param_output	Parameter name in kernel for the output data.
ImpactModel.posterior	Posterior samples by variable name, or None if not set.
ImpactModel.rng_key	Pseudo-random number generator key.
ImpactModel.temp_dir	Temporary directory path, or None if not set.
ImpactModel.vi_result	Variational inference result, or None if not set.

Inference

ImpactModel.train_on_batch(X, y[, rng_key])	Run a single VI step on the given batch of data.
ImpactModel.fit_on_batch(X, y, *[, ...])	Fit the impact model to the provided batch of data.
ImpactModel.fit(X[, y, num_samples, ...])	Fit the impact model to the provided data using epoch-based training.
ImpactModel.set_posterior_sample(...)	Set posterior samples for the model.
ImpactModel.is_fitted()	Check fitted status.
ImpactModel.predict_on_batch(X, *[, ...])	Predict the output based on the fitted model.
ImpactModel.predict(X, *[, intervention, ...])	Predict the output based on the fitted model.
ImpactModel.log_likelihood(X[, y, ...])	Compute the log-likelihood of the data under the given model.

Explicit Sampling

ImpactModel.sample_prior_predictive_on_batch(X, *)	Draw samples from the prior predictive distribution.
ImpactModel.sample_prior_predictive(X, *[, ...])	Draw samples from the prior predictive distribution.
ImpactModel.sample(*[, num_samples, ...])	Draw posterior samples from a fitted model.
ImpactModel.sample_posterior_predictive_on_batch(X, *)	Draw samples from the posterior predictive distribution.
ImpactModel.sample_posterior_predictive(X, *)	Draw samples from the posterior predictive distribution.

Miscellaneous

ImpactModel.estimate_effect([...])	Estimate the effect of an intervention.
ImpactModel.cleanup()	Clean up the temporary directory created for storing outputs.
ImpactModel.cleanup_models()	Clean up temporary directories for all ImpactModel instances.