使用pythonmip库和cvxpy语法

Question

对于混合整数优化问题，我需要使用CBC solver，但是在目标环境中，我不能使用作为外包软件安装的CBC solver，它必须是python库的一部分。为了解决这个问题，我找到了内置CBC解算器附带的mip库https://pypi.org/project/mip/https://docs.python-mip.com/en/latest/index.html，它只能在导入该库后使用，而不需要单独安装CBC解算器。我的问题是，我已经有很多用cvxpy编写的代码（使用这个单独的CBC解算器）。现在的问题是，是否有可能使用CBC内置的mip库，但从常规的cvxpy接口使用它？在不更改代码的情况下，将所有内容重写为mip sytax等

我需要重写为mip语法的示例代码：

import numpy as np
import cvxpy as cp
import cvxopt 
import mip

def run_sample_optimization():
    demand = np.array([[100, 500, 30], [20, 200, 50], [150, 15, 35], [10, 5, 25]])
    product_supply = np.array([550, 200, 170, 40])

    allocation = cp.Variable(demand.shape, integer=True)

    objective = cp.Maximize(cp.sum(allocation/demand))


    constraints =[cp.sum(allocation, axis=1) <= product_supply,
                allocation <= demand,
                allocation >= 0]                        


    problem = cp.Problem(objective, constraints)

    optimal_value = problem.solve(solver=cp.GLPK_MI) # <-- it would be perfect to link somehow from this place to CBC implemented in mip library

    print('product supply:', product_supply)
    print('demand:\n', demand)
    print('allocation:\n', allocation.value)
    print('calculated score:', optimal_value)
    return product_supply, demand, allocation.value, optimal_value

非常感谢

Answer 1

使用此包https://pypi.org/project/mip-cvxpy/ 我用过它，它很管用

This package allows you to solve CVXPY problems using the python-mip package as a backend solver. It works for mixed integer linear problems.

This allows you to use CBC from CVXPY without needing to manually install CBC. By default, CVXOPT calls CyLP to use CBC and requires CBC to be manually installed. python-mip, on the other hand, comes with CBC bundled through pypi.