f3dasm.base package

Submodules

f3dasm.base.data module

class f3dasm.base.data.Data(designspace: f3dasm.base.design.DesignSpace)

Bases: object

Class that contains data

Parameters

data (DataFrame) – data stored in a DataFrame

add(data: pandas.core.frame.DataFrame, ignore_index: bool = False) → None

Add data

Parameters

data (pd.DataFrame) – data to append

add_numpy_arrays(input: numpy.ndarray, output: numpy.ndarray) → None

Append a numpy array to the dataframe

Parameters

  • input (np.ndarray) – 2d numpy array added to input data

  • output (np.ndarray) – 2d numpy array added to output data

add_output(output: numpy.ndarray, label: str) → None

Add a numpy array to the output column of the dataframe

Parameters

  • output (np.ndarray) – Output data

  • label (str) – label of the output column to add to

data: pandas.core.frame.DataFrame
designspace: f3dasm.base.design.DesignSpace
get_input_data() → pandas.core.frame.DataFrame

Get the input data

Returns

DataFrame containing only the input data

Return type

pd.DataFrame

get_n_best_input_parameters_numpy(nosamples: int) → numpy.ndarray

Returns the input vectors in numpy array format of the n best samples

Parameters

nosamples (int) – number of samples

Returns

numpy array containing the n best input parameters

Return type

np.ndarray

get_n_best_output_samples(nosamples: int) → pandas.core.frame.DataFrame

Returns the n lowest rows of the dataframe. Values are compared to the output columns

Parameters

nosamples (int) – number of samples

Returns

DataFrame containing the n best samples

Return type

pd.DataFrame

get_number_of_datapoints() → int

Get the total number of datapoints

Returns

total number of datapoints

Return type

int

get_output_data() → pandas.core.frame.DataFrame

Get the output data

Returns

DataFrame containing only the output data

Return type

pd.DataFrame

plot(input_par1: str, input_par2: Optional[str] = None) → None

Plot the data of two parameters in a figure

Parameters

  • input_par1 (str) – name of first parameter (x-axis)

  • input_par2 (str) – name of second parameter (x-axis)

plot_pairs()

Plot a matrix of 2D plots that visualize the spread of the samples for each dimension. Requires seaborn to be installed.

reset_data() → None

Reset the dataframe to an empty dataframe with the appropriate input and output columns

show() → None

Print the data to the console

f3dasm.base.design module

class f3dasm.base.design.DesignSpace(input_space: List[f3dasm.base.space.ParameterInterface] = <factory>, output_space: List[f3dasm.base.space.ParameterInterface] = <factory>)

Bases: object

Main clas for the design of experiments

Parameters

  • input_space (List[SpaceInterface]) – list of parameters

  • output_space (List[SpaceInterface]) – list of parameters

add_input_space(space: f3dasm.base.space.ParameterInterface) → None

Add a new parameter to the searchspace

Parameters

space (SpaceInterface) – search space parameter to be added

add_output_space(space: f3dasm.base.space.ParameterInterface) → None

Add a new parameter to the searchspace

Parameters

space (SpaceInterface) – search space parameter to be added

get_categorical_input_names() → List[str]

Receive the names of the categorical input parameters

Returns

list of names of categorical input parameters

Return type

List[str]

get_categorical_input_parameters() → List[f3dasm.base.space.CategoricalParameter]

Obtain all the categorical input parameters

Returns

space of categorical input parameters

Return type

List[CategoricalSpace]

get_continuous_input_names() → List[str]

Receive the continuous parameter names of the input space”

Returns

list of names of the continuous input parameters

Return type

List[str]

get_continuous_input_parameters() → List[f3dasm.base.space.ContinuousParameter]

Obtain all the continuous parameters

Returns

space of continuous parameters

Return type

List[ContinuousSpace]

get_discrete_input_names() → List[str]

Receive the names of all the discrete parameters

Returns

list of names

Return type

List[str]

get_discrete_input_parameters() → List[f3dasm.base.space.DiscreteParameter]

Obtain all the discrete parameters

Returns

space of discrete parameters

Return type

List[DiscreteSpace]

get_empty_dataframe() → pandas.core.frame.DataFrame

Create an empty DataFrame with the information of the input and output space

Returns

DataFrame containing “input” and “output” columns

Return type

pd.DataFrame

get_input_names() → List[str]

Get the names of the input parameters

Returns

List of the names of the input parameters

Return type

List[str]

get_input_space() → List[f3dasm.base.space.ParameterInterface]

Get the input space

Returns

List of input parameters

Return type

List[ParameterInterface]

get_number_of_input_parameters() → int

Obtain the number of input parameters

Returns

number of input parameters

Return type

int

get_number_of_output_parameters() → int

Obtain the number of input parameters

Returns

number of output parameters

Return type

int

get_output_names() → List[str]

Get the names of the output parameters

Returns

List of the names of the output parameters

Return type

List[str]

get_output_space() → List[f3dasm.base.space.ParameterInterface]

Get the output space

Returns

List of output parameters

Return type

List[ParameterInterface]

input_space: List[f3dasm.base.space.ParameterInterface]
is_single_objective_continuous() → bool

Check if the output is single objective and continuous

Returns

boolean value if the above described condition is true

Return type

bool

output_space: List[f3dasm.base.space.ParameterInterface]

f3dasm.base.function module

class f3dasm.base.function.Function(noise: bool = False, seed: Optional[Any] = None, dimensionality: int = 2, scale_bounds: numpy.ndarray = array([[0.0, 1.0], [0.0, 1.0]]), input_domain: numpy.ndarray = array([[0.0, 1.0], [0.0, 1.0]]))

Bases: abc.ABC

Interface of a continuous benchmark function

Parameters

  • noise (bool) – inflict Gaussian noise on the output.

  • seed (Any|int) – value to seed the random generator (Default = None).

  • dimensionality (int) – input dimension

  • scale_bounds (Any|List[float]) – array containing the lower and upper bound of the scaling factor of the input data (Default = [0.0, 1.0])

  • input_domain (np.ndarray) – array containing the lower and upper bound of the input domain of the original function (Default = [0.0, 1.0])

check_if_within_bounds(x: numpy.ndarray) → bool

Check if the input vector is between the given scaling bounds

Parameters

x (np.ndarray) – input vector

Returns

whether the vector is within the boundaries

Return type

bool

dfdx(x: numpy.ndarray) → numpy.ndarray

Compute the gradient at a particular point in space. Gradient is computed by numdifftools.

Parameters

x (np.ndarray) – input vector

Returns

gradient

Return type

np.ndarray

dimensionality: int = 2
f(x) → numpy.ndarray

Analytical function of the objective function. Needs to be implemented by inhereted class

input_domain: numpy.ndarray = array([[0., 1.],        [0., 1.]])
noise: bool = False
plot(orientation: str = '3D', px: int = 300, domain: numpy.ndarray = array([[0.0, 1.0], [0.0, 1.0]]), show: bool = True)

Generate a surface plot, either 2D or 3D, of the function

Parameters

  • orientation (str, optional) – Either “2D” or “3D” orientation. Defaults to “3D”.

  • px (int, optional) – Number of points per dimension. Defaults to 300.

  • domain (List, optional) – Domain that needs to be plotted . Defaults to [0, 1].

Returns

Figure and axis

Return type

fig, ax

plot_data(data: f3dasm.base.data.Data, px: int = 300, domain: numpy.ndarray = array([[0.0, 1.0], [0.0, 1.0]]))

Create a 2D contour plot with the datapoints as scatter

Parameters

  • data (Data) – Data object containing samples

  • px (int, optional) – Number of pixels on each axis. Defaults to 300.

  • domain (np.ndarray, optional) – Domain that needs to be plotted. Defaults to np.tile([0.0, 1.0], (2, 1)).

Returns

_description_

Return type

_type_

scale_bounds: numpy.ndarray = array([[0., 1.],        [0., 1.]])
seed: Any = None
static set_seed(seed) → None

Set the numpy seed of the random generator

f3dasm.base.optimization module

class f3dasm.base.optimization.Optimizer(data: f3dasm.base.data.Data, hyperparameters: Optional[Mapping[str, Any]] = <factory>, seed: int = 81320, defaults: Optional[Mapping[str, Any]] = <factory>)

Bases: object

Mainclass to inherit from to implement optimization algorithms

Parameters

  • data (Data) – Data-object

  • hyperparamaters (dict) – Dictionary with hyperparamaters (default is None)

  • seed (int) – seed to set the optimizer

  • defaults (dict) – Default hyperparameter arguments

algorithm: Any
data: f3dasm.base.data.Data
defaults: Optional[Mapping[str, Any]]
extract_data() → f3dasm.base.data.Data

Returns a copy of the data

hyperparameters: Optional[Mapping[str, Any]]
init_parameters() → None

Set the initialization parameters. This could be dynamic or static hyperparameters.

iterate(iterations: int, function: f3dasm.base.function.Function) → None

Calls the update_step function multiple times.

Parameters

  • iterations (int) – number of iterations

  • function (Function) – Objective function to evaluate

seed: int = 81320
set_algorithm() → None

If necessary, the algorithm needs to be set

set_data(data: f3dasm.base.data.Data) → None
static set_seed(seed: int) → None

Set the seed of the optimizer. Needs to be inherited.

update_step(function: f3dasm.base.function.Function) → None

One iteration of the algorithm. Adds the new input vector and resulting output immediately to the data attribute

Parameters

function (Function) – Objective function to evaluate

f3dasm.base.samplingmethod module

class f3dasm.base.samplingmethod.SamplingInterface(doe: f3dasm.base.design.DesignSpace, seed: Optional[Any] = None)

Bases: abc.ABC

Interface for sampling method

Parameters

  • doe (DoE) – design of experiments object

  • seed (Any) – Optional: seed for sampling (default is None)

doe: f3dasm.base.design.DesignSpace
get_samples(numsamples: int) → f3dasm.base.data.Data

Receive samples of the search space

Parameters

numsamples (int) – number of samples

Returns

Data objects with the samples

Return type

Data

sample_continuous(numsamples: int) → numpy.ndarray

Create N samples within the search space

Parameters

  • numsamples (int) – number of samples

  • dimensions (int) – number of dimensions

Returns

samples

Return type

np.array

seed: Any = None
set_seed(seed: int) → None

f3dasm.base.space module

class f3dasm.base.space.CategoricalParameter(name: str, categories: List[str])

Bases: f3dasm.base.space.ParameterInterface

Creates a search space parameter that is categorical

Parameters

categories (list) – list of strings that represent available categories

categories: List[str]
type: str = 'category'
class f3dasm.base.space.ConstraintInterface

Bases: object

Interface for constraints

class f3dasm.base.space.ContinuousParameter(name: str, lower_bound: float = - inf, upper_bound: float = inf)

Bases: f3dasm.base.space.ParameterInterface

Creates a search space parameter that is continuous

Parameters

  • lower_bound (float) – lower bound of continuous search space

  • upper_bound (float) – upper bound of continuous search space (exclusive)

lower_bound: float = -inf
type: str = 'float'
upper_bound: float = inf
class f3dasm.base.space.DiscreteParameter(name: str, lower_bound: int = 0, upper_bound: int = 1)

Bases: f3dasm.base.space.ParameterInterface

Creates a search space parameter that is discrete

Parameters

  • lower_bound (int) – lower bound of discrete search space

  • upper_bound (int) – upper bound of discrete search space (exclusive)

lower_bound: int = 0
type: str = 'int'
upper_bound: int = 1
class f3dasm.base.space.ParameterInterface(name: str)

Bases: object

Interface class of a search space parameter

Parameters

name (str) – name of the parameter

name: str

f3dasm.base.utils module

f3dasm.base.utils.make_nd_continuous_design(bounds: numpy.ndarray, dimensions: int)

Module contents