verticapy.machine_learning.memmodel.decomposition.SVD#

class verticapy.machine_learning.memmodel.decomposition.SVD(vectors: list | ndarray, values: list | ndarray)#

InMemoryModel implementation of the SVD Algorithm.

Parameters#

Note

SVD are defined entirely by their attributes. For example, vectors and values define a SVD model.

Attributes are identical to the input parameters, followed by an underscore (‘_’).

Initalization

Import the required module.

from verticapy.machine_learning.memmodel.decomposition import SVD

A SVD model is defined by its vectors and values. In this example, we will use the following:

vectors = [
    [0.4, 0.5],
    [0.3, 0.2],
]


values = [0.1, 0.3]

Let’s create a SVD model.

model_svd = SVD(vectors, values)

Create a dataset.

data = [[0.3, 0.5]]

Making In-Memory Transformation

Use transform() method to do transformation.

model_svd.transform(data)
Out[6]: array([[2.7       , 0.83333333]])

Deploy SQL Code

Let’s use the following column names:

cnames = ['col1', 'col2']

Use transform_sql() method to get the SQL code needed to deploy the model using its attributes.

model_svd.transform_sql(cnames)
Out[8]: ['col1 * 0.4 / 0.1 + col2 * 0.3 / 0.1', 'col1 * 0.5 / 0.3 + col2 * 0.2 / 0.3']

Use get_attributes() method to check the attributes of the rotated model.

model_svd.get_attributes()
Out[9]: 
{'vectors': array([[0.4, 0.5],
        [0.3, 0.2]]),
 'values': array([0.1, 0.3])}

Hint

This object can be pickled and used in any in-memory environment, just like SKLEARN models.

Methods

`__init__`(vectors, values)
`get_attributes`()	Returns the model attributes.
`set_attributes`(**kwargs)	Sets the model attributes.
`transform`(X)	Transforms and applies the `SVD` model to the input matrix.
`transform_sql`(X)	Transforms and returns the SQL needed to deploy the `SVD` model.

Attributes

Must be overridden in child class