integor 1.2

Optimization and constraint satisfaction for integer linear programs in a user-friendly way

IntegOR

IntegOR is a framework based on scipy.optimize that aims to traduce an Operation Research problem that can be linearized to a binary integer linear problem, to solve it.

The interest is the ability to easily define your variables as strings and to, from here, create easily constraints and cost vector.

Installation

To install the package, you can use pip:

pip install integor

Usage

Suppose you want to solve the following linear problem:

import numpy as np

def J(x):
    """Function to minimize"""
    return x[0] + 2 * x[1] + 3 * x[2] + 4 * x[3]

def f(x):
    """Inequality constraint. f(x) must be positive"""
    return np.array([
        x[0] + x[1] - 1,
        x[2] + x[3] - 1,
        x[0] + x[2] - 1,
    ])

def g(x):
    """Equality constraint. g(x) must be zero"""
    return np.array([
        x[1] + x[3] - 1,
    ])

Import

from integor import set_variable_names, Variable, get_cost_matrix, solve_ilp, get_solution

Set variable names

set_variable_names(["x0", "x1", "x2", "x3"])

Define variables as python objects

x0 = Variable("x0")
x1 = Variable("x1")
x2 = Variable("x2")
x3 = Variable("x3")

Define the constraints from the variables

constraint1 = x0 + x1 >= 1
constraint2 = x2 + x3 >= 1
constraint3 = x0 + x2 >= 1
constraint4 = x1 + x3 == 1
constraints = [constraint1, constraint2, constraint3, constraint4]

Define the cost from the variables

cost_matrix = get_cost_matrix(x0 + 2 * x1 + 3 * x2 + 4 * x3)

Specify integrality ie type of the solution : 0 for continuous, 1 for integer-bounded

integrality = np.ones(len(cost_matrix)) * 1  # You can let this as it is

Solve the problem

# Solve
res = solve_ilp(cost_matrix=cost_matrix, constraints=constraints, integrality=integrality)
solution_vector = res.x

print(res)
print("\nSolution vector: ", solution_vector)

print("\nVerify that the solution satisfies the constraints:")
verify(res.x)

print("\nGet the solution as a dictionary:")
print(get_solution(res.x))