NXTfusion.NXTfusion module

class NXTfusion.NXTfusion.Entity(name, domain, dtype=<class 'numpy.int32'>)

Bases: object

Class representing the Entity concept.

__getitem__(self, x: str)

Method that returns the numeric value internally associated to each object in the Entity class.

Parameters

x (str) – String name of a specific object in the Entity.

Returns

primary key int

__getitem__(self, x: int)

Method that returns the str name of the object with primary key x.

Parameters

x (int) – Primary key (unique id) of an object in the domain represented by the Entity.

Returns

name of the object : str

__init__(name, domain, dtype=<class 'numpy.int32'>)

Constructor for the Entity class.

Parameters

• name (str) – Name of the Entity (use a mnemonic name describing the class of objects represented by the Entity)

• domain (iterable (list) containing str) – List of the possible objects belonging to this class (e.g. patients IDs, proteins Uniprot identifiers, …). It is an unique identifier naming (with a string) all the objects composing the domain of the Entity.

• dtype (np.dtype) – Smallest possible numpy type able to uniquely enumerate all the objects. len(domain) < max_number_representable(dtype).

class NXTfusion.NXTfusion.Relation(name: str, domain1: NXTfusion.NXTfusion.Entity, domain2: NXTfusion.NXTfusion.Entity, data, task: str, loss: torch.nn.modules.module.Module, relationWeight: float, side1=None, side2=None, path=None)

Bases: dict

Class that represent a relation (matrix in MF terms) with all its parameters and functions.

__init__(name: str, domain1: NXTfusion.NXTfusion.Entity, domain2: NXTfusion.NXTfusion.Entity, data, task: str, loss: torch.nn.modules.module.Module, relationWeight: float, side1=None, side2=None, path=None)

Constructor for the Relation class..

Parameters

• name (str) – Mnemonic name of the specific relation/matrix.

• domain1 (Entity) – Entity1 involved in the relation (on dimension 0)

• domain2 (Entity) – Entity2 involved in the relation (on dimension 1)

• data (DataMatrix) – DataMatrix object containing the matrix describing this relation

• task (str ["regression", "binary"]) – Type of prediction task associated to this relation. “Regression” for real valued predictions, “binary” for binary classification.

• loss (NX.NXLosses or t.nn.Module) – Pytorch-like loss module corresponding to the loss that must be used to compute the reconstruction error for this relation.

• relationWeight (float) – A relation-specific weight that will multiply the loss score during training.

class NXTfusion.NXTfusion.MetaRelation(name, domain1, domain2, side1=None, side2=None, relations=[], prediction=False)

Bases: object

Constructor for the Relation class. The Meta Relation represents multi-relations between the same entities (used for example in tensor factorization).

As a convention, we recommend to use names in the form ENT1-ENT2.

The domains must be the same for each relation in it, since the MetaRelation defines a tensor where the dimension 0 and 1 represent the same entities for all the matrices involved in the tensor. Can allow side info (common to all relations in it).

__getitem__(self, x: str)

Getitem method that searches by Relation.name

Parameters

x (str) – The name of a Relation in this MetaRelation

Returns

The target Relation or None

__getitem__(self, x: int)

Getitem method that searches by position of the target Relation in the tensor/MetaRelation.

Parameters

x (str) – The position (index) of a Relation in this MetaRelation

Returns

The target Relation or None

__init__(name, domain1, domain2, side1=None, side2=None, relations=[], prediction=False)

Constructor for the MetaRelation class.

Parameters

• name (str) – Mnemonic name of the specific relation/matrix.

• domain1 (Entity) – Entity1 involved in the relation (on dimension 0)

• domain2 (Entity) – Entity2 involved in the relation (on dimension 1)

• relations (list of NX.Relation objects) – List of the relations involved in this MetaRelation (tensor)

append(r)

Method that adds a Relation object to an existing MetaRelation :param r: :type r: Relation

getPos(x)

next()

pop(pos)

class NXTfusion.NXTfusion.ERgraph(entityList: list, name='')

Bases: list

Class that represents the entire Entity-Relation graphs, namely a list of MetaRelations. Each MetaRelation might contain multiple Relations.

__contains__(self, x: int)

Function that determines whether a specific MetaRelation object is present in the graph.

Parameters

x (MetaRelation) – A MetaRelation object.

Returns

bool (Is x present?)

__contains__(self, x: str)

Function that determines whether a specific MetaRelation.name is present in the graph.

Parameters

x (str) – A MetaRelation str name.

Returns

bool (Is x present?)

__init__(entityList: list, name='')

Constructor for the ERgraph (Entity-Relation Graph) object.

Parameters

entityList (list of MetaRelations) – List of MetaRelations that describe the full Entity Relation graph.

__str__()

Function that expresses the ERgraph as string