CODEWITHSUNDEEP
pythonnumpyoptimizationscipynonlinear-optimization
itssoelif
itssoelif

Reputation: 19

Scipy raise error, TypeError: unsupported operand type(s) for +: 'float' and 'dict' however the variables are float

I'm trying to optimize one constrained, nonlinear model with scipy.

import numpy as np; from scipy.optimize import minimize; import math

# initial guesses
n = 2
x0 = np.zeros(n)
T = 0.1
L = 0.1


def objective(T, L):
    try:
        return (350 / T) + (35 * ((312.5 * (T / 2)) + (11.69 * (math.sqrt(T + L))) + (6.6256 * math.sqrt(T + L))))
    except ValueError:
        return None


def constraint1(T, L):
    (6.67 * math.sqrt(T + L) / (1250 * (T + L))) - 0.02


# show initial objective
print('Initial Objective: ' + str(objective(T, L)))

# optimize
con1 = {'type': 'ineq', 'fun': constraint1}
cons = ([con1])
solution = minimize(objective, x0, args=cons)
x = solution.x

# show final objective
print('Final  Objective: ' + str(objective(T, L)))

# print solution
print('Solution'); print('x1 = ' + str(x[0])); print('x2 = ' + str(x[1]))

When I ran the code, I get the error,

line 24, in objective return (350 / T) + (35 * ((312.5 * (T / 2)) + (11.69 * (math.sqrt(T + L))) + (6.6256 * math.sqrt(T + L)))) TypeError: unsupported operand type(s) for +: 'float' and 'dict'

on the objective(T,L), T described as,

Parameter T of scipyProject.constraint1 T: {truediv, add} pythonProject

but on constraint1(T, L)

Parameter T of scipyProject.constraint1 T: {add} pythonProject

Can you help me, why am I getting this error?

Upvotes: 0

Views: 904

Answers (1)

Deepak H R
Deepak H R

Reputation: 61

I ran your code to debug it. I noticed the following and made the changes accordingly:

  • The function constraint1(T, L) did not return anything.
  • As @ForceBru mentioned, using args=cons will pass the dictionary you made for the constraints as args to your objective function. The scipy documentation defines the objective function to take in a NumPy array as a single argument.
  • The objective function computes math.sqrt(T+L), and there is no constraint mentioning T+L > 0. I got a domain error from the math.sqrt function when I fixed the previous problem. I checked out this answer by @Peter and decided to do something similar using `safeSqrt' function.

Here is your code after some changes. (Btw, I don't know much about this equation, but it looks like setting L = -T is always good. So you can eliminate some terms and optimize for a single variable.)

import numpy as np
from scipy.optimize import minimize
import math

T = 0.1
L = 0.1
x0 = np.array([T, L])


def safeSqrt(s):
    # extend math.sqrt with a safe function
    # ref: https://stackoverflow.com/questions/40372471/math-domain-error-due-to-disrespected-constraint-in-scipy-slsqp-minimize
    return math.sqrt(s) if s > 0 else 0

def objective(x):
    T, L = x[0], x[1]
    linearPart = (312.5 * T / 2)
    sqrtPart = (11.69 + 6.6256) * safeSqrt(T + L)
    return 10 / T + (linearPart + sqrtPart) # your original objective was 35 times this value 

def constraint(x):
    T, L = x[0], x[1]
    result = 6.67 / (1250 * .02) - safeSqrt(T+L) # equivalent to your original constraint
    return result

# show initial objective
print('Initial Objective: ' + str(objective(x0)))
print("contraint: ", constraint(x0))

# optimize
solution = minimize(objective, x0, constraints={"fun": constraint, "type": "ineq"})

x = solution.x

# show final objective
print('Final  Objective: ' + str(objective(x)))

# print solution
print('Solution')
print('x1 = ' + str(x[0]))
print('x2 = ' + str(x[1]))

The code output is:

Initial Objective: 115.625
contraint:  -0.18041359549995795   
Final  Objective: 79.05694150421618
Solution
x1 = 0.25298231697349643
x2 = -1.5173265889148493

Please let me know if this is what you wanted :)

Upvotes: 2

Related Questions