Projection Matrix Factory¶

Module for the factory class for projection matrices

`ProjectionFactory.beta`	Beta distribution
`ProjectionFactory.cauchy`	Cauchy distribution
`ProjectionFactory.dirichlet`	Dirichlet distribution
`ProjectionFactory.laplace`	Laplace distribution
`ProjectionFactory.multivariate_normal`	Multivariate normal distribution
`ProjectionFactory.normal`	Normal distribution
`ProjectionFactory.projections`
`ProjectionFactory.uniform`	Uniform distribution

class ProjectionFactory[source]

Bases: object

Factory class that spawns projection matrices.

Example

>>> from athena import ProjectionFactory
>>> import numpy as np
>>> input_dim = 2
>>> n_features = 40
>>> params = [1.1, 2.9]
>>> for pname in ProjectionFactory.projections:
>>>     y = ProjectionFactory(pname)(input_dim, n_features, params)

static __new__(cls, fname)[source]: Create and return a new object. See help(type) for accurate signature.

static beta(input_dim, n_features, params)[source]

Beta distribution

Parameters

input_dim (int) – dimension of the inputs.
n_features (int) – dimension of the RKHS.
params (list) – the two parameters are the alpha and beta shape parameters respectively.

Returns

n_features-by-input_dim projection matrix.

Return type

numpy.ndarray.

static cauchy(input_dim, n_features, params)[source]

Cauchy distribution

Parameters

input_dim (int) – dimension of the inputs.
n_features (int) – dimension of the RKHS.
params (list) – the single parameter is a scale factor.

Returns

n_features-by-input_dim projection matrix.

Return type

numpy.ndarray.

static dirichlet(input_dim, n_features, params)[source]

Dirichlet distribution

Parameters

input_dim (int) – dimension of the inputs.
n_features (int) – dimension of the RKHS.
params (list) – the single parameter is a scale to the input_dim dimensional shape parameter.

Returns

n_features-by-input_dim projection matrix.

Return type

numpy.ndarray.

static laplace(input_dim, n_features, params)[source]

Laplace distribution

Parameters

input_dim (int) – dimension of the inputs.
n_features (int) – dimension of the RKHS.
params (list) – the single parameter is the scale of the distribution, the mean is set to 0.

Returns

n_features-by-input_dim projection matrix.

Return type

numpy.ndarray.

static multivariate_normal(input_dim, n_features, params)[source]

Multivariate normal distribution

Parameters

input_dim (int) – dimension of the inputs.
n_features (int) – dimension of the RKHS.
params (list) – the input_dim dimensioanl parameters are the

diagonal of the covariance matrix of the distribution. The mean is set to the 0 vector. :return: n_features-by-input_dim projection matrix. :rtype: numpy.ndarray.

static normal(input_dim, n_features, params)[source]

Normal distribution

Parameters

input_dim (int) – dimension of the inputs.
n_features (int) – dimension of the RKHS.
params (list) – the single parameter is the variance of the distribution. The mean is set to 0.

Returns

n_features-by-input_dim projection matrix.

Return type

numpy.ndarray.

projections = ['beta', 'cauchy', 'dirichlet', 'laplace', 'multivariate_normal', 'normal', 'uniform']

static uniform(input_dim, n_features, params)[source]

Uniform distribution

Parameters

input_dim (int) – dimension of the inputs.
n_features (int) – dimension of the RKHS.
params (list) – the two parameters are the extremals of the domain.

Returns

n_features-by-input_dim projection matrix.

Return type

numpy.ndarray.