deephyper.skopt.sampler.sobol.Space

class deephyper.skopt.sampler.sobol.Space(dimensions, model_sdv=None, config_space=None)

Bases: object

Initialize a search space from given specifications.

Parameters:

dimensions (list, shape=(n_dims,)) –

List of search space dimensions. Each search dimension can be defined either as

• a (lower_bound, upper_bound) tuple (for Real or Integer dimensions),

• a (lower_bound, upper_bound, “prior”) tuple (for Real dimensions),

• as a list of categories (for Categorical dimensions), or

• an instance of a Dimension object (Real, Integer or Categorical).

Note

The upper and lower bounds are inclusive for Integer dimensions.

Methods

deactivate_inactive_dimensions	When ConfigSpace is used, it will return the "lower" bound of inactive parameters.
distance	Compute distance between two points in this space.
from_yaml	Create a Space from yaml configuration file.
get_transformer	Returns all transformers as list.
inverse_transform	Inverse transform samples from the warped space back to the original space.
rvs	Draw random samples.
set_transformer	Sets the transformer of all dimension objects to transform.
set_transformer_by_type	Sets the transformer of dim_type objects to transform.
transform	Transform samples from the original space into a warped space.
update_prior	Update the prior of the dimensions.

Attributes

bounds	The dimension bounds, in the original space.
dimension_names	Names of all the dimensions in the search-space.
is_categorical	Space contains exclusively categorical dimensions.
is_partly_categorical	Space contains any categorical dimensions.
is_real	Returns true if all dimensions are Real.
n_constant_dimensions	Returns the number of constant dimensions which have zero degree of freedom, e.g. an Integer dimensions with (0., 0.) as bounds.
n_dims	The dimensionality of the original space.
transformed_bounds	The dimension bounds, in the warped space.
transformed_n_dims	The dimensionality of the warped space.

property bounds

The dimension bounds, in the original space.

deactivate_inactive_dimensions(x)

When ConfigSpace is used, it will return the “lower” bound of inactive parameters.

property dimension_names

Names of all the dimensions in the search-space.

distance(point_a, point_b)

Compute distance between two points in this space.

Args: point_a : array

First point.

point_barray

Second point.

classmethod from_yaml(yml_path, namespace=None)

Create a Space from yaml configuration file.

Parameters:

• yml_path (str) –

  Full path to yaml configuration file, example YaML below: Space:

  • Integer:

  low: -5 high: 5 - Categorical: categories: - a - b - Real: low: 1.0 high: 5.0 prior: log-uniform

• namespace – str, default=None

• use (Namespace within configuration file to)

• first (will use)

• provided (namespace if not)

Returns:

Instantiated Space object.

Return type:

space (Space)

get_transformer()

Returns all transformers as list.

inverse_transform(Xt)

Inverse transform samples from the warped space back to the original space.

Parameters:

Xt (array of floats, shape=(n_samples, transformed_n_dims)) – The samples to inverse transform.

Returns:

The original samples.

Return type:

X (list of lists, shape=(n_samples, n_dims))

property is_categorical

Space contains exclusively categorical dimensions.

property is_partly_categorical

Space contains any categorical dimensions.

property is_real

Returns true if all dimensions are Real.

property n_constant_dimensions

Returns the number of constant dimensions which have zero degree of freedom, e.g. an Integer dimensions with (0., 0.) as bounds.

property n_dims

The dimensionality of the original space.

rvs(n_samples=1, random_state=None, n_jobs=1)

Draw random samples.

The samples are in the original space. They need to be transformed before being passed to a model or minimizer by space.transform().

Parameters:

• n_samples (int, default=1) – Number of samples to be drawn from the space.

• random_state (int or RandomState, Optional) – Set random state to something other than None for reproducible results.

• n_jobs (int) – the number of parallel processes to use to perform sampling.

Returns:

Points sampled from the space.

Return type:

points (list of lists, shape=(n_points, n_dims))

set_transformer(transform)

Sets the transformer of all dimension objects to transform.

Parameters:

transform (str or list of str) – Sets all transformer,, when transform is a string. Otherwise, transform must be a list with strings with the same length as dimensions

set_transformer_by_type(transform, dim_type)

Sets the transformer of dim_type objects to transform.

Parameters:

• transform (str) – Sets all transformer of type dim_type to transform

• dim_type (type) – Can be deephyper.skopt.space.Real, deephyper.skopt.space.Integer or deephyper.skopt.space.Categorical.

transform(X)

Transform samples from the original space into a warped space.

Note: this transformation is expected to be used to project samples

into a suitable space for numerical optimization.

Parameters:

X (list of lists, shape=(n_samples, n_dims)) – The samples to transform.

Returns:

The transformed samples.

Return type:

Xt (array of floats, shape=(n_samples, transformed_n_dims))

property transformed_bounds

The dimension bounds, in the warped space.

property transformed_n_dims

The dimensionality of the warped space.

update_prior(X, y, q=0.9)

Update the prior of the dimensions.

Instead of doing random-sampling, a kernel density estimation is fit on the region of interest and sampling is performed from this distribution.