Contents
Is it possible to optimization an absolute value function?
Absolute value functions themselves are very difficult to perform standard optimization procedures on. They are not continuously differentiable functions, are nonlinear, and are relatively difficult to operate on.
Can a model be reformulated with absolute values?
Absolute values in the objective function. Absolute values as part of the objective function of a model can also be reformulated to become linear, in certain cases. If the objective is a minimization problem of the form or is a maximization problem of the form , then the model can easily be reformulated to be solved using linear programming.
How to perform linear programming with absolute values?
This function is effectively the combination two piecewise functions: if and if . This methodology is the basis of performing linear programming with absolute values. In the case , the expression can be reformulated as and . This relation is easiest to see using a number line, as follows:
How to minimize the sum of absolute deviations?
Minimizing the sum of absolute deviations. Let deviations be represented by , where i is the observation, gives the deviation, is an observation. To minimize the deviation, the problem is formulated in a basic form as: as the objective function, and linear constraints are.
Which is the constraint for the objective function?
That is, we know that L ≥ 0, W ≥ 0, and H ≥ 0. This constraint can be used to reduce the number of variables in the objective function, V = LWH, from three to two. We can choose to solve the constraint for any convenient variable, so let’s solve it for H .
When do we have constraints in an optimization problem?
Anytime we have a closed region or have constraints in an optimization problem the process we’ll use to solve it is called constrained optimization . In this section we will explore how to use what we’ve already learned to solve constrained optimization problems in two ways.
Which is the relation between absolute values in constraints?
Absolute values in constraints. This relation is easiest to see using a number line, as follows: Figure 1: Number line depicting the above absolute value problem. This number line represents both the absolute value function as well as the two combined linear functions described above, demonstrating that the two formulations are equivalent.