将 'senses' 和 'rhs' 从 CPLEX 转换为 CVXPY

Question

我正在尝试使用 CPLEX 求解器将一些代码从纯 CPLEX 转换为 CVXPY。原始代码是这样的：

    p = cplex.Cplex()
    p.objective.set_sense(p.objective.sense.maximize)

    obj = np.zeros(numCols)

    for i in range(numSamples):
        if labels[i] == 0:
            for b in range(numBuckets):
                obj[zv_ + b*numSamples + i] = C
        else:
            for b in range(numBuckets):
                obj[zv_ + numBuckets*numSamples + b*numSamples + i] = 1


    p.linear_constraints.add(rhs=rhs,senses=senses)
    p.linear_constraints.set_names(zip(range(constraint_cnt),cnames))


    lb = np.zeros(numCols)
    ub = np.ones(numCols)
    p.variables.add(obj = obj, lb = lb, ub = ub, columns=cols, types=types, names=names)

    p.parameters.timelimit.set(maxtime)
    p.solve()
    sol = p.solution

这是我将其转换为 CVXPY 语法的尝试：

    import cvxpy as cp
    obj = np.zeros(numCols)

    for i in range(numSamples):
        if labels[i] == 0:
            for b in range(numBuckets):
                obj[zv_ + b*numSamples + i] = C
        else:
            for b in range(numBuckets):
                obj[zv_ + numBuckets*numSamples + b*numSamples + i] = 1

    objective = cp.Maximize(sum(obj))

令我非常困惑的是 CPLEX 指定 'rhs' 以及 'senses' 的方式。这些应该是什么？

Answer 1

Cplex.linear_constraints.add 的 CPLEX Python API 文档说：

senses must be either a list of single-character strings or a string containing the senses of the linear constraints. Each entry must be one of 'G', 'L', 'E', and 'R', indicating greater-than, less-than, equality, and ranged constraints, respectively.

rhs is a list of floats, specifying the righthand side of each linear constraint.

CVXPY 是一种建模语言，因此与其将意义指定为 'G'、'L'、'E'，不如使用 >=、<= , 和 ==（即 Python 比较运算符）。约束的右侧将是单个约束的数字（例如 Python float 或 int），或者可能是一组约束的 numpy 数组。

CVXPY examples 应该能让您很好地了解如何做事。

将 'senses' 和 'rhs' 从 CPLEX 转换为 CVXPY

Converting 'senses' and 'rhs' from CPLEX to CVXPY

cplex

cvxpy