surfaces 0.0.4

A collection and visualization of black-box objective functions

Surfaces

A collection and visualization of single objective black-box functions for optimization benchmarking

Visualizations

Objective Function	Heatmap	Surface Plot
Sphere function
Rastrigin function
Ackley function
Rosenbrock function
Beale function
Himmelblaus function
Hölder Table function
Cross-In-Tray function

Installation

The most recent version of Surfaces is available on PyPi:

pip install surfaces

Example

import numpy as np

from surfaces.convex import SphereFunction
from surfaces.non_convex import AckleyFunction
from surfaces.visualize import plotly_surface


sphere_function = SphereFunction(n_dim=2, metric="score")
ackley_function = AckleyFunction(metric="loss")


step_ = 0.05
min_ = 10
max_ = 10
search_space = {
    "x0": np.arange(-min_, max_, step_),
    "x1": np.arange(-min_, max_, step_),
}

plotly_surface(sphere_function, search_space).show()
plotly_surface(ackley_function, search_space).show()

API reference

Objective Function Classes

All objective function classes have the following parameters:

• metric: "score" or "loss"
• input_type: "dictionary" or "arrays"

Each objective function class has the following parameters:

SphereFunction

• A = 1

AckleyFunction

• A = 20
• B = 2 * pi

RastriginFunction

• A = 10
• B = 2 * pi

RosenbrockFunction

• A = 1
• B = 100

BealeFunction

• A = 1.5
• B = 2.25
• C = 2.652

HimmelblausFunction

• A = -11
• B = -7

HölderTableFunction

• A = 10
• B = 1

CrossInTrayFunction

• A = -0.0001
• B = 100
• C = 1