AlgorithmRegistry¶

from causaliq_discovery import AlgorithmRegistry

The central registry that maps (algorithm, variant) pairs to AlgorithmSpec instances and PackageAdapter classes. All built-in algorithm specs are pre-registered at import time. Adapters are registered separately when their implementation modules are imported.

AlgorithmRegistry¶

Class methods¶

algorithms()¶

AlgorithmRegistry.algorithms() -> List[str]

Return a sorted list of all registered algorithm names.

>>> AlgorithmRegistry.algorithms()
['gs', 'h2pc', 'hc', 'hc-stable', 'iiamb', 'mmhc', 'pc-stable', 'tabu', 'tabu-stable']

variants(algorithm)¶

AlgorithmRegistry.variants(algorithm: str) -> List[str]

Return a sorted list of variant names for the given algorithm.

>>> AlgorithmRegistry.variants("hc")
['bnlearn', 'causaliq']

Raises ValueError if the algorithm is not registered.

get_spec(algorithm, variant)¶

AlgorithmRegistry.get_spec(
    algorithm: str,
    variant: Optional[str],
) -> AlgorithmSpec

Return the AlgorithmSpec for the given algorithm/variant pair. When variant is None, the default variant (first registered) is used.

Raises ValueError if the algorithm or variant is not registered.

get_adapter(algorithm, variant)¶

AlgorithmRegistry.get_adapter(
    algorithm: str,
    variant: Optional[str],
) -> Type[PackageAdapter]

Return the adapter class for the given algorithm/variant pair.

Raises ValueError if algorithm or variant is not registered. Raises NotImplementedError if no adapter has been registered for this variant yet.

register_spec(spec)¶

AlgorithmRegistry.register_spec(spec: AlgorithmSpec) -> None

Register an AlgorithmSpec. Overwrites any existing spec for the same (algorithm, variant) key.

register_adapter(algorithm, variant, adapter_class)¶

AlgorithmRegistry.register_adapter(
    algorithm: str,
    variant: str,
    adapter_class: Type[PackageAdapter],
) -> None

Associate a PackageAdapter subclass with a registered spec.

Raises ValueError if the (algorithm, variant) pair has no registered spec.

AlgorithmSpec¶

Dataclass capturing the full specification for one algorithm variant.

@dataclass
class AlgorithmSpec:
    algorithm: str
    variant: str
    package: str
    description: str
    graph_type: str
    supported_hyperparameters: Set[str] = ...
    hyperparameter_defaults: Dict[str, Any] = ...
    hyperparameter_name_map: Dict[str, str] = ...
    hyperparameter_value_map: Dict[str, Dict[Any, Any]] = ...
    paper_ref: str = ""
    paper_url: str = ""
    algorithm_class: str = ""

Attributes¶

Attribute	Type	Description
`algorithm`	`str`	Common algorithm name, e.g. `"hc"`.
`variant`	`str`	Variant name, e.g. `"causaliq"` or `"bnlearn"`.
`package`	`str`	Package providing the implementation, e.g. `"CausalIQ"` or `"bnlearn"`.
`description`	`str`	Human-readable description of the algorithm.
`graph_type`	`str`	Type of graph produced, e.g. `"DAG"` or `"CPDAG"`.
`supported_hyperparameters`	`Set[str]`	Common hyperparameter names accepted by this variant.
`hyperparameter_defaults`	`Dict[str, Any]`	Default value for each supported hyperparameter, keyed by common name.
`hyperparameter_name_map`	`Dict[str, str]`	Mapping from common hyperparameter name to the package-specific argument name. Omitted entries use the common name unchanged.
`hyperparameter_value_map`	`Dict[str, Dict]`	Mapping from common hyperparameter value to the package-specific value, keyed by common name then common value.
`paper_ref`	`str`	Bibliographic reference for the original paper.
`paper_url`	`str`	URL to the original paper.
`algorithm_class`	`str`	Algorithmic class: `"score"`, `"constraint"`, or `"hybrid"`.

Property¶

supported_hyperparameters¶

Derived set of hyperparameter names supported by this spec. Used by learn_graph for validation.

HyperparameterSpec¶

Dataclass documenting a single hyperparameter. Used by the CLI describe command to display help.

@dataclass
class HyperparameterSpec:
    name: str
    category: str
    type: str
    description: str
    valid_values: Optional[List[str]] = None
    default_display: Optional[str] = None

Attributes¶

Attribute	Type	Description
`name`	`str`	Common hyperparameter name, e.g. `"score"`.
`category`	`str`	Logical group: `"score"`, `"constraint"`, or `"general"`.
`type`	`str`	Python type name: `"int"`, `"float"`, or `"str"`.
`description`	`str`	Short one-line description.
`valid_values`	`List[str] \\| None`	Enumerated valid values, or `None` if any value of the declared type is accepted.
`default_display`	`str \\| None`	Display string for the default shown in `cqdisc describe`, e.g. `"No limit"`.