Optimization cost function file

Is it possible that optimization can be done without creating a seperate function file for the cost function and the constraint?..if yes then how can we do it....?

[Rik]

What function are you using? Most functions will allow you to have a function handle as an input, so you can include a handle to a function in that same m-file (or even use an anonymous function).

[Aquatris]

Check the document page of "fmincon" where they have examples for this. Most solvers for Matlab follow similar input format.

[Mondeep maz]

It is an objective function consisting a sum of variables...problem is that i want to design the optimization such that no of variables of the objective function at each iteration can get changed...if i use a function handle i wont b able to change it i suppose...so i want a way to get around this problem

[sabreen farouq]

where i can find this famous cost function ?

[Walter Roberson]

There are a few example cost functions such as rosenbeck, but mostly you write your own cost function depending on what you want to optimize.

See also Problem Based Optimization

Answer 1

[Walter Roberson]

Each iteration of the objective function called by fmincon or ga or similar must involve the same number of variables, and must be deterministic (no random values.)

You can use different numbers of variables for different calls to the minimizer, including in a loop. That could involve creating a new anonymous function for each iteration of the loop.

An objective function which is just the sum of variables can be expressed as just @sum; for a weighted sum, @(x) sum(x.*weights) or more robustly @(x) sum(x(:).*weights(:)) to be sure you do not have row/column difficulties.