neuralogic.core.constructs package

Subpackages

neuralogic.core.constructs.function package

Submodules

neuralogic.core.constructs.factories module

class AtomFactory[source]

Bases: object

Factory for creating atoms (relations) in the logic program. It supports dot notation for predicate names and provides access to special and hidden factories.

get(name: str) → BaseRelation[source]

Creates a relation with the given name and 0 arity.

Parameters:: name (str) – The name of the predicate.
Returns:: The created relation.
Return type:: BaseRelation

static get_predicate(name: str, arity: int, hidden: bool, special: bool) → Predicate[source]

class ConstantFactory[source]

Bases: object

Factory for creating constants. Constants are automatically converted to lowercase unless specified otherwise.

get(item: str, const_type: str | None = None) → Constant[source]

Creates a constant with the given name and optional type.

Parameters:

item (str) – The name of the constant.
const_type (str, optional) – The type of the constant. Default: None.

Returns:

The created constant.

Return type:

Constant

class HiddenPredicateFactory[source]

Bases: object

Factory for creating hidden predicates. Hidden predicates are not part of the output unless explicitly requested.

get(name: str) → BaseRelation[source]

property special: SpecialPredicateFactory

class SpecialPredicateFactory(hidden: bool = False)[source]

Bases: object

Factory for creating special predicates, such as ‘alldiff’, ‘neq’, ‘eq’, etc. Special predicates are handled differently by the backend engine.

add(a: Any, b: Any, c: Any) → BaseRelation[source]

add_eval(a: Any, b: Any) → BaseRelation[source]

alldiff(*args: Any) → BaseRelation[source]

anypred() → BaseRelation[source]

div_eval(a: Any, b: Any) → BaseRelation[source]

eq(a: Any, b: Any) → BaseRelation[source]

geq(a: Any, b: Any) → BaseRelation[source]

get(name: str) → BaseRelation[source]

Creates a special relation with the given name and 0 arity.

Parameters:: name (str) – The name of the special predicate.
Returns:: The created special relation.
Return type:: BaseRelation

gt(a: Any, b: Any) → BaseRelation[source]

property hidden: SpecialPredicateFactory

leq(a: Any, b: Any) → BaseRelation[source]

lt(a: Any, b: Any) → BaseRelation[source]

max_eval(a: Any, b: Any) → BaseRelation[source]

maxcard(*args: Any) → BaseRelation[source]

min_eval(a: Any, b: Any) → BaseRelation[source]

mod(a: Any, b: Any, c: Any) → BaseRelation[source]

mod_eval(a: Any, b: Any) → BaseRelation[source]

mul_eval(a: Any, b: Any) → BaseRelation[source]

neq(a: Any, b: Any) → BaseRelation[source]

next(a: Any, b: Any) → BaseRelation[source]

sub(a: Any, b: Any, c: Any) → BaseRelation[source]

sub_eval(a: Any, b: Any) → BaseRelation[source]

truepred() → BaseRelation[source]

class VariableFactory[source]

Bases: object

Factory for creating variables. Variables are automatically capitalized unless specified otherwise.

get(item: str, var_type: str | None = None) → Variable[source]

Creates a variable with the given name and optional type.

Parameters:

item (str) – The name of the variable.
var_type (str, optional) – The type of the variable. Default: None.

Returns:

The created variable.

Return type:

Variable

neuralogic.core.constructs.java_objects module

class JavaFactory(settings: SettingsProxy | None = None)[source]

Bases: object

Factory for converting high-level Python logic constructs (Atoms, Rules, etc.) into their corresponding Java objects. It maintains internal factories for predicates, constants, and weights.

add_metadata_function(metadata: Metadata, map: Any, function: str) → None[source]

add_parametrized_function(metadata: Metadata, metadata_obj: Any, function: str) → None[source]

atom_to_clause(atom: Any) → Any[source]

Converts a single atom to a Java Clause.

Parameters:: atom (Any) – The atom to convert.
Returns:: The Java Clause object.
Return type:: Any

get_conjunction(relations: Iterable[Any], variable_factory: Any, default_weight: Any = None, is_example: bool = False) → Any[source]

get_generic_relation(relation_class: Any, relation: Any, variable_factory: Any, default_weight: Any = None, is_example: bool = False) → Any[source]

Generic method to convert a Python relation to a Java relation object.

Parameters:

relation_class (Any) – The Java class of the relation (e.g., BodyAtom, ValuedFact).
relation (Any) – The Python relation object.
variable_factory (Any) – The Java variable factory.
default_weight (Any, optional) – A default weight if the relation is unweighted. Default: None.
is_example (bool) – Whether the relation is part of an example. Default: False.

Returns:

The Java relation object.

Return type:

Any

get_lifted_example(example: Any, learnable_facts: bool = False) → tuple[Any, Any][source]

Converts a Python example to a Java LiftedExample.

Parameters:

example (Any) – The example to convert.
learnable_facts (bool) – Whether facts in the example should be learnable. Default: False.

Returns:

A tuple containing the Java label conjunction and the Java LiftedExample.

Return type:

tuple[Any, Any]

get_metadata(metadata: Metadata | None, metadata_class: Any) → Any[source]

get_new_weight_factory() → Any[source]

get_predicate(predicate: Any) → Any[source]

get_predicate_metadata_pair(predicate_metadata: Any) → Any[source]

get_query(query: Any) → tuple[Any, list[Any] | None][source]

Converts a Python query to a Java representation.

Parameters:: query (Any) – The query to convert. Can be a Rule, list of relations, or a single relation.
Returns:: A tuple containing the Java head relation and a list of Java body relations.
Return type:: tuple[Any, list[Any] | None]

get_relation(relation: Any, variable_factory: Any, is_example: bool = False) → Any[source]

get_rule(rule: Any) → Any[source]

Converts a Python Rule to a Java WeightedRule.

Parameters:: rule (Any) – The Python rule to convert.
Returns:: The Java WeightedRule object.
Return type:: Any

get_term(term: Any, variable_factory: Any) → Any[source]

Converts a Python term (Variable, Constant, str, int, float) to a Java term.

Parameters:

term (Any) – The term to convert.
variable_factory (Any) – The Java variable factory to use for variables.

Returns:

The Java term object.

Return type:

Any

get_valued_fact(relation: Any, variable_factory: Any, default_weight: Any = None, is_example: bool = False) → Any[source]

get_variable_factory() → Any[source]

get_weight(weight: Any, name: str | None, fixed: bool) → Any[source]

to_clause(atoms: Iterable[Any]) → Any[source]

class ValueFactory[source]

Bases: object

Factory for converting between Python values (scalars, lists, numpy arrays) and Java Value objects. It handles ScalarValue, VectorValue, and MatrixValue.

static from_java(value: Any) → float | list | ndarray[source]

Converts a Java Value object to a Python representation (float, list, or numpy array).

Parameters:: value (Any) – The Java Value object to convert.
Returns:: The Python representation of the value.
Return type:: float | list | np.ndarray

get_value(weight: Any) → tuple[bool, Any][source]

Converts a Python weight to a Java Value object.

Parameters:: weight (Any) – The weight to convert. Can be a scalar, tuple (for dimensions), list, or numpy array.
Returns:: A tuple containing a boolean (True if initialized, False if just dimensions) and the Java Value object.
Return type:: tuple[bool, Any]

neuralogic.core.constructs.metadata module

Bases: object

Represents metadata for a logic construct (e.g., rule, predicate).

Metadata can specify properties like learnability, transformation functions, aggregation functions, and combination functions.

aggregation

combination

combine(other: Metadata) → Metadata[source]

Combines this metadata with another Metadata object. Values from the other object take precedence.

Parameters:: other (Metadata) – The other Metadata object to combine with.
Returns:: A new combined Metadata object.
Return type:: Metadata

copy() → Metadata[source]

Returns a shallow copy of the metadata.

Returns:: The copy of the metadata.
Return type:: Metadata

duplicate_grounding

static from_iterable(iterable: Iterable[Any]) → Metadata[source]

Creates a Metadata object from an iterable of functions or values.

Parameters:: iterable (Iterable) – The iterable containing metadata entries.
Returns:: The created Metadata object.
Return type:: Metadata

learnable

transformation

neuralogic.core.constructs.predicate module

class Predicate(name: str, arity: int, hidden: bool = False, special: bool = False)[source]

Bases: object

Represents a predicate in the logic program, defined by its name and arity.

arity

hidden

name

set_arity(arity: int) → Predicate | None[source]

special

to_str() → str[source]

Returns a string representation of the predicate (without arity).

Returns:: The string representation.
Return type:: str

class PredicateMetadata(predicate: Predicate, metadata: Metadata)[source]

Bases: object

Associates metadata with a predicate.

metadata

predicate

neuralogic.core.constructs.relation module

class BaseRelation(predicate: Predicate, terms: Any = None, function: TransformationFunction | CombinationFunction | None = None, negated: bool = False)[source]

Bases: object

Represents a relation with a predicate, terms, and an optional activation function.

property T: BaseRelation

attach_activation_function(function: TransformationFunction | CombinationFunction) → BaseRelation[source]

Attaches an activation or combination function to the relation.

Parameters:: function (Union[TransformationFunction, CombinationFunction]) – The function to attach.
Returns:: A new relation with the attached function.
Return type:: BaseRelation

function

negated

predicate

terms

to_str(end=False) → str[source]

Returns a string representation of the relation.

Parameters:: end (bool) – Whether to append a dot at the end. Default: False.
Returns:: The string representation.
Return type:: str

class WeightedRelation(weight: Any, predicate: Predicate, fixed: bool = False, terms: Any = None, function: TransformationFunction | CombinationFunction | None = None)[source]

Bases: BaseRelation

Represents a relation with an associated weight (learnable or fixed).

property T: WeightedRelation

attach_activation_function(function: Transformation | Combination) → WeightedRelation[source]

Attaches an activation or combination function to the relation.

Parameters:: function (Union[TransformationFunction, CombinationFunction]) – The function to attach.
Returns:: A new relation with the attached function.
Return type:: BaseRelation

fixed() → WeightedRelation[source]

Returns a copy of the relation with the weight fixed.

Returns:: The weighted relation with a fixed weight.
Return type:: WeightedRelation

is_fixed

to_str(end=False)[source]

Returns a string representation of the relation.

Parameters:: end (bool) – Whether to append a dot at the end. Default: False.
Returns:: The string representation.
Return type:: str

weight

weight_name

neuralogic.core.constructs.rule module

class Rule(head: BaseRelation, body: RuleBody | Iterable[BodyItem] | BodyItem)[source]

Bases: object

Represents a rule in the model, consisting of a head and a body.

body: list[BodyItem] | FContainer

head

metadata: Metadata | None

to_str(_: bool = False) → str[source]

class RuleBody(lit1: BaseRelation, lit2: BaseRelation)[source]

Bases: object

Represents the body of a rule, which is a collection of literals.

literals: list['BaseRelation' | FContainer]

metadata: Metadata | None

neuralogic.core.constructs.term module

class Constant(name: str, type: str | None = None)[source]

Bases: object

Represents a constant in a logic expression.

name

type

class Variable(name: str, type: str | None = None)[source]

Bases: object

Represents a variable in a logic expression.

name

type

neuralogic.core.constructs package

Subpackages

Submodules

neuralogic.core.constructs.factories module

neuralogic.core.constructs.java_objects module

neuralogic.core.constructs.metadata module

neuralogic.core.constructs.predicate module

neuralogic.core.constructs.relation module

neuralogic.core.constructs.rule module

neuralogic.core.constructs.term module

Module contents