Module datasets

Synthetic dataset.

operalib.datasets.toy_data_curl_free_field(n_samples, loc=25, space=0.5)

Curl-Free toy dataset.

Generate a scalar field as mixture of five gaussians at location:

• (0 , 0)
• (0 , loc)
• ( loc, 0)
• (-loc, 0)
• (0 , -loc)

whith variance equal to ‘space’. Then return the gradient of the field. The return result is a pair (inputs, targets) of arrays.

Parameters:

n_samples : int

Number of samples to generate.

loc: float, optional (default = 25.)

Centers of the Gaussians.

space: float, optional (default = .5)

Variance of the Gaussians.

Returns:

X : array, shape = [n_samples, 2]

Array of evenly space points.

y : array shape = [n_samples, 2]

Array corresponding to the velocity at the coordinates present in inputs.

See also

operalib.toy_data_curl_free_mesh

Generate Curl-Free mesh.

operalib.toy_data_div_free_mesh

Generate Divergence-Free mesh.

operalib.datasets.toy_data_div_free_mesh(n_samples, loc=25.0, space=0.5)

Divergence-Free toy dataset.

Generate a scalar field as mixture of five gaussians at location:

• (0 , 0)
• (0 , loc)
• ( loc, 0)
• (-loc, 0)
• (0 , -loc)

whith variance equal to ‘space’. Then return the orthogonal of gradient of the field. The return result is a pair of meshes.

Parameters:

n_points : int

Number of samples to generate.

loc: float, optional (default = 25.)

Centers of the Gaussians.

space: float, optional (default = .5)

Variance of the Gaussians.

Returns:

X, Y : :rtype: (array, array), shape = [n, n]

Mesh, X, Y coordinates.

U, V : :rtype: (array, array), shape = [n, n]

Mesh, (U, V) velocity at (X, Y) coordinates

See also

operalib.toy_data_curl_free_field

Generate Curl-Free field.

operalib.toy_data_div_free_field

Generate Divergence-Free field.

operalib.datasets.toy_data_div_free_field(n_samples, loc=25, space=0.5)

Divergence-Free toy dataset.

Generate a scalar field as mixture of five gaussians at location:

• (0 , 0)
• (0 , loc)
• ( loc, 0)
• (-loc, 0)
• (0 , -loc)

whith variance equal to ‘space’. Then return the orthogonal of gradient of the field. The return result is a pair (inputs, targets) of arrays.

Parameters:

n_points : int

Number of samples to generate.

loc: float, optional (default = 25.)

Centers of the Gaussians.

space: float, optional (default = .5)

Variance of the Gaussians.

Returns:

X : array, shape = [n_samples, 2]

Array of evenly space points.

y : array shape = [n_samples, 2]

Array corresponding to the velocity at the coordinates present in inputs.

See also

operalib.toy_data_div_free_mesh

Generate Curl-Free mesh.

operalib.toy_data_div_free_mesh

Generate Divergence-Free mesh.

operalib.datasets.toy_data_curl_free_mesh(n_samples, loc=25.0, space=0.5)

Curl-Free toy dataset.

Generate a scalar field as mixture of five gaussians at location:

• (0 , 0)
• (0 , loc)
• ( loc, 0)
• (-loc, 0)
• (0 , -loc)

whith variance equal to ‘space’. Then return the gradient of the field. The return result is a pair meshes.

Parameters:

n_samples : int

Number of samples to generate.

loc: float, optional (default = 25.)

Centers of the Gaussians.

space: float, optional (default = .5)

Variance of the Gaussians.

Returns:

X, Y : :rtype: (array, array), shape = [n, n]

Mesh, X, Y coordinates.

U, V : :rtype: (array, array), shape = [n, n]

Mesh, (U, V) velocity at (X, Y) coordinates

See also

operalib.toy_data_curl_free_field

Generate Curl-Free field.

operalib.toy_data_div_free_field

Generate Divergence-Free field.

operalib.datasets.array2mesh(X, side=None)

Array to mesh converter.

Parameters:

X : array, shape = [n_samples, 2]

The inputs array.

Returns:

x_mesh : array, shape = [n_samples, n_samples]

The x_axis of the mesh corresponding to inputs.

y_mesh : array, shape = [n_samples, n_samples]

The y_axis of the mesh corresponding to inputs

operalib.datasets.mesh2array(x_mesh, y_mesh)

Mesh to array converter.

Parameters:

x_mesh : array, shape = [n_samples, n_samples]

The x_axis mesh.

y_mesh : array, shape = [n_samples, n_samples]

The y_axis mesh.

Returns:

inputs : array, shape = [n_samples, 2]

The inputs corresponding to the mesh (x_mesh, y_mesh).

operalib.datasets.toy_data_quantile(n_samples=50, probs=0.5, pattern=SinePattern(enveloppe=(0.3333333333333333, 1.0), inputs_bound=(0.0, 1.5), random_state=<mtrand.RandomState object>, sine_period=1.0), noise=(1.0, 0.2, 0.0, 1.5), random_state=None)

Sine wave toy dataset.

The target y is computed as a sine curve at of modulated by a sine envelope with some period (default 1/3Hz) and mean (default 1). Moreover, this pattern is distorted with a random Gaussian noise with mean 0 and a linearly decreasing standard deviation (default from 1.2 at X = 0 to 0.2 at X = 1 . 5).

Parameters:

n_samples : int

Number of samples to generate.

probs : list or float, shape = [n_quantiles], default=0.5

Probabilities (quantiles levels).

pattern : callable, default = SinePattern()

Callable which generates n_sample 1D data (inputs and targets).

noise : :rtype: (float, float, float, float)

Noise parameters (variance, shift, support_min, support_max).

Returns:

X : array, shape = [n_samples, 1]

Input data.

y : array, shape = [n_sample, 1]

Targets.

quantiles : array, shape = [n_samples x n_quantiles]

True conditional quantiles.