了解scipy优化中的约束

Question

I am trying to understand constraints in the scipy optimize function. I want to minimize the function under the assumption, that input values would always be positive. So I have my constraint function defined as follows:

def apply_constraint(inputs):
    return inputs[0] - inputs[0]

What would happen if all my inputs are negative? I tried with negative inputs but could not understand the result. What is the constraint attribute really meant for? It also gives results with negative values.

Here is my complete code.

from __future__ import division
import numpy as np
from scipy.optimize import minimize


def f(x):
    Y = ((x + 100) / 100)
    return Y[0]


def apply_constraint(inputs):
    return inputs[0] - inputs[0]


my_constraints = ({'type': 'eq', "fun": apply_constraint})

min_result = minimize(f, [2], method="SLSQP", options={
                    'disp': False}, bounds=[(-2, 101)], constraints=my_constraints)

print("Minima found at: X = {}, Y = {}".format(min_result.x, min_result.fun))

Answer 1

Your constraint function is incorrect

inputs[0] - inputs[0] will always be 0, it should be

def apply_constraint(inputs):
    return inputs[0]

Constraints work in one of two ways either type = 'eq' or type = 'ineq'

The constraint function will be evaluated and compared with 0

For an equality constraint the constraint function will be forced to = 0

For an inequality constraint the constraint function will be forced to be non-negative ie >=0

In this case using the constraint function above and type = 'ineq' should work