## Global Optimization Toolbox |

This example shows how to create and minimize a fitness function using the Genetic Algorithm in the Global Optimization Toolbox.

On this page… |
---|

A Fitness Function with Additional Arguments |

Here we want to minimize a simple function of two variables

min f(x) = 100 * (x(1)^2 - x(2)) ^2 + (1 - x(1))^2; x

We create a MATLAB file named simple_fitness.m with the following code in it:

function y = simple_fitness(x) y = 100 * (x(1)^2 - x(2)) ^2 + (1 - x(1))^2;

The Genetic Algorithm solver assumes the fitness function will take one input x where x is a row vector with as many elements as number of variables in the problem. The fitness function computes the value of the function and returns that scalar value in its one return argument y.

To minimize our fitness function using the `ga` function, we need to pass in a function handle to the fitness function as well as specifying the number of variables in the problem.

FitnessFunction = @simple_fitness; numberOfVariables = 2; [x,fval] = ga(FitnessFunction,numberOfVariables)

Optimization terminated: average change in the fitness value less than options.TolFun. x = 0.4725 0.2136 fval = 0.2876

The x returned by the solver is the best point in the final population computed by `ga`. The fval is the value of the function @simple_fitness evaluated at the point x.

**A Fitness Function with Additional Arguments**

Sometimes we want our fitness function to be parameterized by extra arguments that act as constants during the optimization. For example, in the previous fitness function, say we want to replace the constants 100 and 1 with parameters that we can change to create a family of objective functions. We can re-write the above function to take two additional parameters to give the new minimization problem

min f(x) = a * (x(1)^2 - x(2)) ^2 + (b - x(1))^2; x

a and b are parameters to the fitness function that act as constants during the optimization (they are not varied as part of the minimization). One can create a MATLAB file called parameterized_fitness.m containing the following code:

function y = parameterized_fitness(x,a,b) y = a * (x(1)^2 - x(2)) ^2 + (b - x(1))^2;

**Minimizing Using Additional Arguments**

Again, we need to pass in a function handle to the fitness function as well as the number of variables as the second argument.

`ga` will call our fitness function with just one argument 'x', but our fitness function has three arguments: x, a, b. We can use an anonymous function to capture the values of the additional arguments, the constants a and b. We create a function handle 'FitnessFunction' to an anonymous function that takes one input 'x', but calls 'parameterized_fitness' with x, a, and b. The variables a and b have values when the function handle 'FitnessFunction' is created, so these values are captured by the anonymous function.

```
a = 100; b = 1; % define constant values
FitnessFunction = @(x) parameterized_fitness(x,a,b);
numberOfVariables = 2;
[x,fval] = ga(FitnessFunction,numberOfVariables)
```

Optimization terminated: maximum number of generations exceeded. x = 0.8331 0.6835 fval = 0.0390

**Vectorizing Your Fitness Function**

Consider the previous fitness function again:

f(x) = a * (x(1)^2 - x(2)) ^2 + (b - x(1))^2;

By default, the `ga` solver only passes in one point at a time to the fitness function. However, sometimes speed up can be achieved if the fitness function is vectorized to take a set of points and return a set of function values.

For example if the solver wants to evaluate a set of five points in one call to this fitness function, then it will call the function with a matrix of size 5-by-2, i.e. , 5 rows and 2 columns (recall 2 is the number of variables).

Create a MATLAB file called vectorized_fitness.m with the following code:

function y = vectorized_fitness(x,a,b) y = a * (x(:,1).^2 - x(:,2)).^2 + (b - x(:,1)).^2;

This vectorized version of the fitness function takes a matrix x with an arbitrary number of points, the rows of x, and returns a column vector y with the same number of rows as x.

We need to specify that the fitness function is vectorized using the options structure created using `gaoptimset`. The options structure is passed in as the ninth argument.

FitnessFunction = @(x) vectorized_fitness(x,100,1); numberOfVariables = 2; options = gaoptimset('Vectorized','on'); [x,fval] = ga(FitnessFunction,numberOfVariables,[],[],[],[],[],[],[],options)

Optimization terminated: average change in the fitness value less than options.TolFun. x = -0.4862 0.2451 fval = 2.2163