NXTfusion.DataMatrix module¶
-
class
NXTfusion.DataMatrix.
DataMatrix
[source]¶ Bases:
object
The input “data” format should be: {(ent1, ent2): value} for all the observed elements in the matrix.
The format in which the data is stored in the DataMatrix object is the following: featsHT = {domain1Name_numeric : [ numpy16_domain2Names_numeric, numpyX_labels ]}
-
__init__
(self, name: str, ent1: Entity, ent2: Entity, data: numpy.ndarray)¶ One of the alternative constructors for the DataMatrix class.
- Parameters
name (str) – Name of the data matrix
ent1 (Entity) – Entity object representing the object on the dimension 0
ent2 (Entity) – Entity object representing the object on the dimension 1
data (dict) – Hash table containing the (sparse) elements and in the matrix describing the relation. The input “data” format should be: {(ent1, ent2): value} for all the observed elements in the matrix.
dtype (numpy.dtype) – The smallest possible type that could be used to store the elements of the matrix (e.g. np.int16 can represent up to 2^16 unique objects in the entity)
-
__init__
(self, name: str, ent1: NX.Entity, ent2: NX.Entity, data: dict, dtype: type)¶ One of the alternative constructors for the DataMatrix class.
- Parameters
name (str) – Name of the data matrix
ent1 (Entity) – Entity object representing the object on the dimension 0
ent2 (Entity) – Entity object representing the object on the dimension 1
data (numpy.ndarray) – Numpy matrix containing the (dense) describing the relation between ent1 and en2.
dtype (numpy.dtype) – The smallest possible type that could be used to store the elements of the matrix (e.g. np.int16)
- Returns
the message id
-
__init__
(self, name: str, data: numpy.ndarray, dtype: numpy.dtype)¶ Simplified constructor for the DataMatrix class. Entities are inferred from the dimensionality of the np.ndarray.
- Parameters
name (str) – Name of the data matrix
data (numpy.ndarray) – Numpy matrix containing the (dense) describing the relation between ent1 and en2.
dtype (numpy.dtype) – The smallest possible type that could be used to store the elements of the matrix (e.g. np.int16 can represent up to 2^16 unique objects in the entity)
-
__init__
(self, path: str)¶ Constructor that reads the DataMatrix from a previously serialized DataMatrix object.
- Parameters
path (str) – Path of the serialized DataMatrix
-
size
()[source]¶ Function that return the size of the relation (number of elements in the matrix).
- Returns
- Return type
Size of the relation in the DataMatrix object
-
-
class
NXTfusion.DataMatrix.
SideInfo
[source]¶ Bases:
object
Class that encapsulated the side information raw data in order to be efficiently processed by NXTfusion. You can use this class to wrap side information vectors analogously to how DataMatrix wraps matrix/relations.
-
__init__
(self, name: str, ent1: Entity, ent2: Entity, data: dict)¶ One of the alternative constructors for the SideInfo class.
- Parameters
name (str) – Name of the data matrix
ent1 (Entity) – Entity object representing the object on the dimension 0
data (dict) – Dict containing ent1 objects as keys and feature vectors (side information) as values.
-
__init__
(self, name: str, ent1: Entity, ent2: Entity, data: numpy.ndarray)¶ One of the alternative constructors for the SideInfo class.
- Parameters
name (str) – Name of the data matrix
ent1 (Entity) – Entity object representing the object on the dimension 0
data (numpy.ndarray) – Numpy array that contains the side information. It has shape (ent1 obj, feature length), similarly to a scikit-learn feature vector.
-
__init__
(self, name: str, ent1: Entity, ent2: Entity, data: scipy.sparse.coo_matrix)¶ One of the alternative constructors for the SideInfo class.
- Parameters
name (str) – Name of the data matrix
ent1 (Entity) – Entity object representing the object on the dimension 0
data (scipy.sparse.coo_matrix) – Scipy coo_matrix that contains the side information. It has shape (ent1 obj, feature length), similarly to a scikit-learn feature vector. It can be sparse, but currently the sparsity during mini batching is NOT supported.
-
__init__
(self, path: str)¶ This constructor reads a serialized (SideInfo.dump()) SideInfo object. :param str path: Path to the serialized SideInfo object.
-