torch_concepts.nn.AncestralSamplingInference

class AncestralSamplingInference(probabilistic_model: ProbabilisticModel, graph_learner: BaseGraphLearner | None = None, log_probs: bool = True, **dist_kwargs)

Ancestral sampling inference for probabilistic graphical models.

This inference engine performs ancestral (forward) sampling by drawing samples from the distributions defined by each variable. It’s useful for generating realistic samples from the model and for tasks requiring stochastic predictions.

The sampling respects the probabilistic structure: - Samples from Bernoulli distributions using .sample() - Uses reparameterization (.rsample()) for RelaxedBernoulli and RelaxedOneHotCategorical - Supports custom distribution kwargs (e.g., temperature for Gumbel-Softmax)

Parameters:

• probabilistic_model – The probabilistic model to perform inference on.

• graph_learner – Optional graph learner for weighted adjacency structure.

• **dist_kwargs – Additional kwargs passed to distribution constructors (e.g., temperature for relaxed distributions).

Example

>>> import torch
>>> from torch.distributions import Bernoulli
>>> from torch_concepts import InputVariable
>>> from torch_concepts.distributions import Delta
>>> from torch_concepts.nn import AncestralSamplingInference, ParametricCPD, ProbabilisticModel
>>> from torch_concepts import EndogenousVariable
>>> from torch_concepts.nn import LinearCC
>>>
>>> # Create a simple PGM: embedding -> A -> B
>>> embedding_var = InputVariable('embedding', parents=[], distribution=Delta, size=10)
>>> var_A = EndogenousVariable('A', parents=['embedding'], distribution=Bernoulli, size=1)
>>> var_B = EndogenousVariable('B', parents=['A'], distribution=Bernoulli, size=1)
>>>
>>> # Define CPDs
>>> from torch.nn import Identity, Linear
>>> cpd_emb = ParametricCPD('embedding', parametrization=Identity())
>>> cpd_A = ParametricCPD('A', parametrization=Linear(10, 1))
>>> cpd_B = ParametricCPD('B', parametrization=LinearCC(1, 1))
>>>
>>> # Create probabilistic model
>>> pgm = ProbabilisticModel(
...     variables=[embedding_var, var_A, var_B],
...     parametric_cpds=[cpd_emb, cpd_A, cpd_B]
... )
>>>
>>> # Create ancestral sampling inference engine
>>> inference = AncestralSamplingInference(pgm)
>>>
>>> # Perform inference - returns samples, not endogenous
>>> x = torch.randn(4, 10)  # batch_size=4, embedding_size=10
>>> results = inference.predict({'embedding': x})
>>>
>>> # Results contain binary samples {0, 1} for Bernoulli variables
>>> print(results['A'].shape)  # torch.Size([4, 1])
>>> print(results['A'].unique())  # tensor([0., 1.]) - actual samples
>>> print(results['B'].shape)  # torch.Size([4, 1])
>>> print(results['B'].unique())  # tensor([0., 1.]) - actual samples
>>>
>>> # Query specific concepts - returns concatenated samples
>>> samples = inference.query(['B', 'A'], evidence={'embedding': x})
>>> print(samples.shape)  # torch.Size([4, 2])
>>> # samples contains [sample_B, sample_A] for each instance
>>> print(samples)  # All values are 0 or 1
>>>
>>> # Multiple runs produce different samples (stochastic)
>>> samples1 = inference.query(['A'], evidence={'embedding': x})
>>> samples2 = inference.query(['A'], evidence={'embedding': x})
>>> print(torch.equal(samples1, samples2))  # Usually False (different samples)
>>>
>>> # With relaxed distributions (requires temperature)
>>> from torch.distributions import RelaxedBernoulli
>>> var_A_relaxed = InputVariable('A', parents=['embedding'],
...                               distribution=RelaxedBernoulli, size=1)
>>> pgm = ProbabilisticModel(
...     variables=[embedding_var, var_A_relaxed, var_B],
...     parametric_cpds=[cpd_emb, cpd_A, cpd_B]
... )
>>> inference_relaxed = AncestralSamplingInference(pgm, temperature=0.05)
>>> # Now uses reparameterization trick (.rsample())
>>>
>>> # Query returns continuous values in [0, 1] for relaxed distributions
>>> relaxed_samples = inference_relaxed.query(['A'], evidence={'embedding': x})
>>> # relaxed_samples will be continuous, not binary

__init__(probabilistic_model: ProbabilisticModel, graph_learner: BaseGraphLearner | None = None, log_probs: bool = True, **dist_kwargs)

Initialize the inference module.

Methods

__init__(probabilistic_model[, ...])	Initialize the inference module.
add_module(name, module)	Add a child module to the current module.
apply(fn)	Apply fn recursively to every submodule (as returned by .children()) as well as self.
bfloat16()	Casts all floating point parameters and buffers to bfloat16 datatype.
buffers([recurse])	Return an iterator over module buffers.
children()	Return an iterator over immediate children modules.
compile(*args, **kwargs)	Compile this Module's forward using torch.compile().
cpu()	Move all model parameters and buffers to the CPU.
cuda([device])	Move all model parameters and buffers to the GPU.
double()	Casts all floating point parameters and buffers to double datatype.
eval()	Set the module in evaluation mode.
extra_repr()	Return the extra representation of the module.
float()	Casts all floating point parameters and buffers to float datatype.
forward(x, *args, **kwargs)	Forward pass delegates to the query method.
get_buffer(target)	Return the buffer given by target if it exists, otherwise throw an error.
get_extra_state()	Return any extra state to include in the module's state_dict.
get_parameter(target)	Return the parameter given by target if it exists, otherwise throw an error.
get_parent_kwargs(parametric_cpd[, ...])	Prepare keyword arguments for CPD forward pass based on parent outputs.
get_results(results, parent_variable)	Sample from the distribution parameterized by the results.
get_submodule(target)	Return the submodule given by target if it exists, otherwise throw an error.
half()	Casts all floating point parameters and buffers to half datatype.
ipu([device])	Move all model parameters and buffers to the IPU.
load_state_dict(state_dict[, strict, assign])	Copy parameters and buffers from state_dict into this module and its descendants.
modules()	Return an iterator over all modules in the network.
mtia([device])	Move all model parameters and buffers to the MTIA.
named_buffers([prefix, recurse, ...])	Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.
named_children()	Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.
named_modules([memo, prefix, remove_duplicate])	Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.
named_parameters([prefix, recurse, ...])	Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.
parameters([recurse])	Return an iterator over module parameters.
predict(external_inputs[, debug, device])	Perform forward pass prediction across the entire probabilistic model.
query(query_concepts, evidence[, debug, device])	Execute forward pass and return only specified concepts concatenated.
register_backward_hook(hook)	Register a backward hook on the module.
register_buffer(name, tensor[, persistent])	Add a buffer to the module.
register_forward_hook(hook, *[, prepend, ...])	Register a forward hook on the module.
register_forward_pre_hook(hook, *[, ...])	Register a forward pre-hook on the module.
register_full_backward_hook(hook[, prepend])	Register a backward hook on the module.
register_full_backward_pre_hook(hook[, prepend])	Register a backward pre-hook on the module.
register_load_state_dict_post_hook(hook)	Register a post-hook to be run after module's load_state_dict() is called.
register_load_state_dict_pre_hook(hook)	Register a pre-hook to be run before module's load_state_dict() is called.
register_module(name, module)	Alias for add_module().
register_parameter(name, param)	Add a parameter to the module.
register_state_dict_post_hook(hook)	Register a post-hook for the state_dict() method.
register_state_dict_pre_hook(hook)	Register a pre-hook for the state_dict() method.
requires_grad_([requires_grad])	Change if autograd should record operations on parameters in this module.
set_extra_state(state)	Set extra state contained in the loaded state_dict.
set_submodule(target, module[, strict])	Set the submodule given by target if it exists, otherwise throw an error.
share_memory()	See torch.Tensor.share_memory_().
state_dict(*args[, destination, prefix, ...])	Return a dictionary containing references to the whole state of the module.
to(*args, **kwargs)	Move and/or cast the parameters and buffers.
to_empty(*, device[, recurse])	Move the parameters and buffers to the specified device without copying storage.
train([mode])	Set the module in training mode.
type(dst_type)	Casts all parameters and buffers to dst_type.
unrolled_probabilistic_model()	Build an 'unrolled' view of the ProbabilisticModel based on graph_learner adjacency.
xpu([device])	Move all model parameters and buffers to the XPU.
zero_grad([set_to_none])	Reset gradients of all model parameters.

Attributes

T_destination
available_query_vars	Get all variable names available for querying.
call_super_init
dump_patches
training