optirustic 0.4.0

No project description provided

Optirustic Py

This is a Python package that mainly let users import serialised data from JSON files exported with the optirustic crate.

It lets you:

• import data into Python classes for easy manipulation;
• calculate the population hyper-volume;
• plot 2D, 3D or parallel coordinate charts of the Pareto front.
• initialise the configurations of genetic operators and algorithms from Python in case someone wants to create a Python library that wraps this library along with a specific optimisation problem.

NOTE: this pacage will not let you optimise a problem from Python, but a dedicated Python wrapper can be create in Rust from a pre-defined problem.

Installation

The package can be installed from PyPi:

pip install optirustic

Usage

These are two example scripts to fetch data and plot the Pareto fronts of the optirustic example files

Python API

All Python API are available in your editor via type hints:

from optirustic import NSGA3

# Load the NSGA3 data first
data = NSGA3(r"../examples/results/DTLZ1_3obj_NSGA3_gen400.json")

# Fetch the problem data
p = data.problem
print(p.number_of_variables)
print(p.variables)
# Fetch the lower bound of X!
print(p.variables["x1"].min_value)

# Get the objective stored into the problem
print(p.objectives)
# Fetch the direction of objective f1
print(p.objectives["f1"].direction)

# Fetch the problem constraints
print(p.constraints)

# Fetch other data such as the algorithm name or generation
print(f"Algorithm name: {data.algorithm}")
print(f"Population reached generation: {data.generation}")
print(f"Algorithm took: {data.took}")
print(f"JSOn file exported on: {data.exported_on}")

# Fetch data for the first individual
print(data.individuals[0])
print(data.individuals[0].constraint_violation)
print(f"Objective values: {data.individuals[0].objectives}")
print(f"Objective f2 value is: {data.individuals[0].get_objective_value("f2")}")
print(f"Variable values: {data.individuals[0].variables}")
print(f"Additional stored data: {data.individuals[0].data}")

# Calculate the hyper-volume
print(f"Hyper-volume is: {data.hyper_volume(reference_point=[100, 100, 100])}")

Generate Pareto front chart

import matplotlib.pyplot as plt
from optirustic import NSGA2, NSGA3

# Plot a 2D charts for a 2-objective problem
NSGA2(r"../examples/results/SCH_2obj_NSGA2_gen250.json").plot()
plt.show()

# Plot a 3D charts for a 3-objective problem
NSGA3(r"../examples/results/DTLZ1_3obj_NSGA3_gen400.json").plot()
plt.show()

# Plot a parallel coordinate chart for an 8-objective problem
NSGA3(r"../examples/results/DTLZ1_8obj_NSGA3_gen750.json").plot()
plt.show()

Generate convergence chart

This template script plots the algorithm convergence by calculating the hyper-volumes at different generations:

import matplotlib.pyplot as plt
from optirustic import NSGA2

# provide the folder where optimistic exported the JSON files
# and a reference point to use in the hyper-volume calculation
NSGA2.plot_convergence(
    folder=r"../examples/results/convergence",
    reference_point=[10000, 10000]
)
plt.show()

Generate reference points

To generate, plot and inspect the reference points for the NSGA3 algorithm you can us:

One layer

import matplotlib.pyplot as plt
from optirustic import DasDarren1998

ds = DasDarren1998(number_of_objectives=3, number_of_partitions=5)
points = ds.calculate()
ds.plot(points)
plt.show()

Two layers

import matplotlib.pyplot as plt
from optirustic import DasDarren1998

two_layers = dict(
    boundary_layer=3,
    inner_layer=4,
    scaling=None,
)
ds = DasDarren1998(number_of_objectives=3, number_of_partitions=two_layers)
points = ds.calculate()
ds.plot(points)
plt.show()

Initialise an algorithm properties

from optirustic import (
    StoppingConditionValue,
    StoppingCondition,
    Algorithm,
    NSGA2Arg,
)

# set the stopping condition at 300 seconds
cond = StoppingConditionValue.max_duration(300)
print(cond)

# define the NSGA2 arguments
args = NSGA2Arg(
    number_of_individuals=10,
    stopping_condition=StoppingCondition(condition=cond),
)
print(args)

# define the algorithm to use in a Python wrapper
algo = Algorithm.nsga2(args)
print(algo)