Ansary–Panda (AP)

This family comprises the AP1, AP2, AP3, and AP4 constructors. All four problems are drawn from “A modified Quasi-Newton method for vector optimization problem” [2].

Overview

The table below shows the default dimensions and componentwise variable bounds for each constructor.

ProblemnvarnobjLower boundUpper bound
AP123-10.010.0
AP212-100.0100.0
AP322-100.0100.0
AP433-10.010.0

Analytical Jacobians and Hessians are registered for all four constructors. The catalog metadata classifies the strict convexity of the objective functions as follows:

  • AP1: [:not_strictly_convex, :strictly_convex, :strictly_convex]
  • AP2: [:strictly_convex, :strictly_convex]
  • AP3: [:not_strictly_convex, :not_strictly_convex]
  • AP4: [:not_strictly_convex, :strictly_convex, :strictly_convex]

Mathematical formulations

The formulas below describe the objective functions implemented by the constructors. For each problem, let $F:\mathbb{R}^n \to \mathbb{R}^m$ be defined by $F(x)=(f_1(x),\ldots,f_m(x))$.

AP1

The objectives are

\[\begin{aligned} f_1(x) &= \frac{1}{4} \left( (x_1 - 1)^4 + 2(x_2 - 2)^4 \right), \\ f_2(x) &= \exp\left( \frac{x_1 + x_2}{2} \right) + x_1^2 + x_2^2, \\ f_3(x) &= \frac{1}{6} \left( \exp(-x_1) + 2\exp(-x_2) \right). \end{aligned}\]

where $x = (x_1, x_2) \in \mathbb{R}^2$.

AP2

The objectives are

\[\begin{aligned} f_1(x) &= x_1^2 - 4, \\ f_2(x) &= (x_1 - 1)^2. \end{aligned}\]

where $x \in \mathbb{R}^1$.

AP3

The objectives are

\[\begin{aligned} f_1(x) &= \frac{1}{4} \left( (x_1 - 1)^4 + 2(x_2 - 2)^4 \right), \\ f_2(x) &= (x_2 - x_1^2)^2 + (1 - x_1)^2. \end{aligned}\]

where $x = (x_1, x_2) \in \mathbb{R}^2$.

AP4

The objectives are

\[\begin{aligned} f_1(x) &= \frac{1}{9} \left( (x_1 - 1)^4 + 2(x_2 - 2)^4 + 3(x_3 - 3)^4 \right), \\ f_2(x) &= \exp\left( \frac{x_1 + x_2 + x_3}{3} \right) + x_1^2 + x_2^2 + x_3^2, \\ f_3(x) &= \frac{1}{12} \left( 3\exp(-x_1) + 4\exp(-x_2) + 3\exp(-x_3) \right). \end{aligned}\]

where $x = (x_1, x_2, x_3) \in \mathbb{R}^3$.

Usage

using MOProblems

prob = AP1()
x = [0.0, 0.0]

values = eval_f(prob, x)
J = eval_jacobian(prob, x)
H = eval_hessian(prob, x)

Constructor reference

MOProblems.AP1Function
AP1()

Construct the fixed two-variable, three-objective AP1 problem.

The variables are bounded in [-10, 10]^2. An analytical Jacobian and objective Hessians are registered.

source
MOProblems.AP2Function
AP2()

Construct the fixed one-variable, two-objective AP2 problem.

The variable is bounded in [-100, 100]. An analytical Jacobian and objective Hessians are registered.

source
MOProblems.AP3Function
AP3()

Construct the fixed two-variable, two-objective AP3 problem.

The variables are bounded in [-100, 100]^2. An analytical Jacobian and objective Hessians are registered.

source
MOProblems.AP4Function
AP4()

Construct the fixed three-variable, three-objective AP4 problem.

The variables are bounded in [-10, 10]^3. An analytical Jacobian and objective Hessians are registered.

source