scimba_torch.integration.monte_carlo_parameters¶

Parameter samplers for Monte Carlo simulations.

Classes

`ParametricSampler`(bounds, **kwargs)	Abstract base class for parametric samplers.
`SobolParametricSampler`(bounds, **kwargs)	Sample from a Sobol sequence within given bounds for each dimension.
`UniformParametricSampler`(bounds, **kwargs)	Sample uniformly from given bounds for each dimension.
`UniformVelocitySampler`(velocity_domain)	Sample uniformly the velocities in a circle of radius r.
`UniformVelocitySamplerOnCuboid`(velocity_domain)	Sample uniformly the velocities in a cuboid.

class ParametricSampler(bounds, **kwargs)[source]¶

Bases: ABC

Abstract base class for parametric samplers.

Parameters:

bounds (list[tuple[float, float]]) – A list of tuples, containing the lower and upper bounds.
**kwargs – Keyword arguments including: - pre-sampling: whether to pre-sample points before training; - n_pre_sampled: the number of pre-sampled points to generate if so.

Raises:

TypeError –
- If parameters domain is not a list of tuples of two floats. - If n_pre_sampled is not an integer.
ValueError – If any bound has lower value greater than upper value.

set_new_bounds(nbounds)[source]¶

Updates the bounds of the parameters in the approximation space.

Parameters:: nbounds (list[tuple[float, float]]) – A list of tuples containing the new bounds for each parameter.
Return type:: None

abstract sample_new_points(n)[source]¶

Generates samples within the specified bounds for each dimension.

Parameters:: n (int) – The number of samples to generate.
Return type:: LabelTensor
Returns:: A tensor containing the generated samples and corresponding labels.

sample_from_pre_sampled_points(n)[source]¶

Samples from the pre-sampled points.

Parameters:: n (int) – The number of samples to generate.
Return type:: LabelTensor
Returns:: A tensor containing the generated samples and corresponding labels.

check_sample_size(n)[source]¶

Checks if the sample size is a non-negative integer.

Parameters:

n (int) – The number of samples to generate.

Raises:

TypeError – If argument is not an integer.
ValueError – If argument is negative.

rescale_samples(samples)[source]¶

Rescales the samples to fit within the specified bounds.

This function assumes that the samples are generated in the unit hypercube \([0, 1]^{\text{dim}}\) and rescales them to fit within the bounds specified for each dimension. It also: - make the samples require gradients; - converts the samples to a LabelTensor with zero labels.

Parameters:: samples (Tensor) – A tensor containing the generated samples.
Return type:: LabelTensor
Returns:: A tensor containing the rescaled samples.

class UniformParametricSampler(bounds, **kwargs)[source]¶

Bases: ParametricSampler

Sample uniformly from given bounds for each dimension.

Parameters:: bounds (list[tuple[float, float]]) – A list of tuples, containing the lower and upper bounds.

sample_new_points(n)[source]¶

Generates samples uniformly within the specified bounds for each dimension.

Parameters:: n (int) – The number of samples to generate.
Return type:: LabelTensor
Returns:: A tensor containing the generated samples and corresponding labels.

class SobolParametricSampler(bounds, **kwargs)[source]¶

Bases: ParametricSampler

Sample from a Sobol sequence within given bounds for each dimension.

Parameters:

bounds (list[tuple[float, float]]) – A list of tuples, containing the lower and upper bounds.
**kwargs – Keyword arguments including: - pre-sampling: whether to pre-sample points before training; - n_pre_sampled: the number of pre-sampled points to generate if so.

Raises:

TypeError – If bounds is not a list of tuples of two floats.

sample_new_points(n)[source]¶

Generates samples from a Sobol sequence within the specified bounds.

Parameters:: n (int) – The number of samples to generate.
Return type:: LabelTensor
Returns:: A tensor containing the generated samples and corresponding labels.

class UniformVelocitySampler(velocity_domain)[source]¶

Bases: object

Sample uniformly the velocities in a circle of radius r.

Parameters:: velocity_domain (Circle2D) – Velocity domain in which the velocity will be drawn.
Raises:: TypeError – If velocity domain is not an object of class Circle2D.

dim¶: The number of dimensions, here equal to 2.

sample(n)[source]¶

Generates samples uniformly within the specified bounds for each dimension.

Parameters:

n (int) – The number of samples to generate.

Return type:

LabelTensor

Returns:

A tensor containing the generated samples and corresponding labels.

Raises:

TypeError – If argument is not an integer.
ValueError – If argument is negative.

class UniformVelocitySamplerOnCuboid(velocity_domain)[source]¶

Bases: object

Sample uniformly the velocities in a cuboid.

Parameters:: velocity_domain (Segment1D | Square2D | Cube3D) – Velocity domain in which the velocity will be drawn.
Raises:: TypeError – If velocity domain is not an object of class Segment1D, Square2D or Cube3D.

dim¶: The number of dimensions.

domain_size¶: The size of the domain in each dimension.

lower_bound¶: The lower bound of the domain.

sample(n)[source]¶

Generates samples uniformly within the specified bounds for each dimension.

Parameters:

n (int) – The number of samples to generate.

Return type:

LabelTensor

Returns:

A tensor containing the generated samples and corresponding labels.

Raises:

TypeError – If argument is not an integer.
ValueError – If argument is negative.