使用scipy.optimize.minimize时输入权重向量形状异常问题求助

• 核心错误：目标函数参数顺序不匹配
  scipy.optimize.minimize的执行逻辑是：将优化变量（即x0传入的权重向量）作为第一个参数传递给目标函数，之后才是args中指定的额外参数。你当前的ES_calc_sum函数把pi_bad放在第一个参数位，导致参数传递完全错位：

  • 原本的权重向量被传给了pi_bad参数
  • 原本的pi_bad标量被传给了weights参数
    这就是为什么打印weights时显示单个数值，后续矩阵乘法x @ cov_good_matrix因输入维度不匹配报错。

• 修复步骤

1. 调整目标函数ES_calc_sum的参数顺序，将weights放在第一位：

def ES_calc_sum(weights, pi_bad, mu_good_vector, mu_bad_vector, cov_good_matrix, cov_bad_matrix, alpha):
    print(weights)  # 现在会正确打印权重向量
    sum_val = np.sum(risk_contribution(pi_bad, weights, mu_good_vector, mu_bad_vector, cov_good_matrix, cov_bad_matrix, alpha))
    return sum_val

注意：避免使用sum作为变量名，它是Python内置函数，容易引发冲突

2. 同步修正minimize调用中的args参数顺序，确保和目标函数的参数对应：

res = minimize(ES_calc_sum, 
               x0=weights, 
               args=(pi_bad, mu_good_vector, mu_bad_vector, cov_good_matrix, cov_bad_matrix, alpha), 
               method="SLSQP", 
               bounds=bnds, 
               constraints=cont, 
               tol=0.0001)

3. 修复约束条件中的语法错误（多余的嵌套括号）：
   原约束中第二个等式的fun存在多余括号，修正后：

cont = ({"type": "eq", "fun": lambda x: np.sum(x) - 1},
        {"type": "eq", "fun": lambda x: (1 - pi_bad) * np.sqrt(x @ cov_good_matrix @ x) + pi_bad * np.sqrt(x @ cov_bad_matrix @ x) - target_vol})

• 完整修正后的代码

def ES_calc_sum(weights, pi_bad, mu_good_vector, mu_bad_vector, cov_good_matrix, cov_bad_matrix, alpha):
    print(weights)
    sum_val = np.sum(risk_contribution(pi_bad, weights, mu_good_vector, mu_bad_vector, cov_good_matrix, cov_bad_matrix, alpha))
    return sum_val

weights = np.array([0.5,0.25,0.25])

def RC_minimizer(target_vol, weights, pi_bad, mu_good_vector, 
    mu_bad_vector, cov_good_matrix, cov_bad_matrix, alpha):

     bnds = Bounds(0, 1)
     cont = ({"type": "eq", "fun": lambda x: np.sum(x) - 1},
            {"type": "eq", "fun": lambda x: (1 - pi_bad) * np.sqrt(x @ cov_good_matrix @ x) + pi_bad * np.sqrt(x @ cov_bad_matrix @ x) - target_vol})

     res = minimize(ES_calc_sum, x0=weights, args=(pi_bad, mu_good_vector, mu_bad_vector, cov_good_matrix, cov_bad_matrix, alpha), method="SLSQP", bounds=bnds, constraints=cont, tol=0.0001)

     w_opt = res.x

     return w_opt

内容的提问来源于stack exchange，提问作者Ginner