SVIProcess

class dynode.infer.inference.SVIProcess(*, numpyro_model: ~typing.Callable[[~dynode.config.simulation_config.SimulationConfig, ~typing.Tuple[~jax.Array, ...] | ~jax.Array | None], ~diffrax._solution.Solution], inference_prngkey: ~jax.Array = Array([      0, 8675314], dtype=uint32), num_iterations: ~typing.Annotated[int, ~annotated_types.Gt(gt=0)], num_samples: ~typing.Annotated[int, ~annotated_types.Gt(gt=0)], guide_class: ~typing.Type[~numpyro.infer.autoguide.AutoContinuous] = <class 'numpyro.infer.autoguide.AutoMultivariateNormal'>, guide_init_strategy: ~typing.Callable = <function init_to_median>, optimizer: ~numpyro.optim._NumPyroOptim = <factory>, progress_bar: bool = True, guide_kwargs: dict = <factory>)

Bases: InferenceProcess

Inference process for fitting a numpyro_model to data using SVI.

__init__(**data: Any) → None

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

Methods

get_samples([_, exclude_deterministic])	Get the posterior samples from the inference process.
infer(**kwargs)	Fit the numpyro_model to data using SVI.
model_post_init(context, /)	This function is meant to behave like a BaseModel method to initialise private attributes.
to_arviz()	Return the results of a fit as an arviz InferenceData object.

Attributes

model_config	Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].
num_iterations
num_samples
guide_class
guide_init_strategy
optimizer
progress_bar
guide_kwargs
numpyro_model
inference_prngkey

model_config: ClassVar[ConfigDict] = {'arbitrary_types_allowed': True}

Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

num_iterations: Annotated[int, Gt(gt=0)]

num_samples: Annotated[int, Gt(gt=0)]

guide_class: Type[AutoContinuous]

guide_init_strategy: Callable

optimizer: _NumPyroOptim

progress_bar: bool

guide_kwargs: dict

infer(**kwargs) → SVI

Fit the numpyro_model to data using SVI.

Additional keyword arguments are passed to the numpyro_model.

Returns

SVI

The SVI object used for inference.

get_samples(_: bool = False, exclude_deterministic: bool = True) → dict[str, Array]

Get the posterior samples from the inference process.

Parameters

_bool

Unused parameter, whether or not to group posterior samples by chain or not. SVI does not have chains so this is unnecessary.

exclude_deterministicbool

whether or not to exclude parameters generated from numpyro.deterministic as keys in the returned dictionary, by default True.

Returns

dict[str, Array]

A dictionary of posterior samples, where keys are parameter sites and values are the corresponding samples.

Notes

Keep in mind that posterior samples are generated after the fitting process for SVI, and the samples are not arranged by chain/sample like in MCMC.

to_arviz() → InferenceData

Return the results of a fit as an arviz InferenceData object.

Returns

arviz.InferenceData

arviz InferenceData object containing both priors and posterior_predictive.

Raises

AssertionError

if fitting has not yet been run via infer()

_abc_impl = <_abc._abc_data object>

model_post_init(context: Any, /) → None

This function is meant to behave like a BaseModel method to initialise private attributes.

It takes context as an argument since that’s what pydantic-core passes when calling it.

Args:

self: The BaseModel instance. context: The context.