jppbsi / libopt Goto Github PK

For many problems one needs to optimize vectors or integers (from N1 to N2). I tried with many methods by casting the variables to integer e.g. (int) a->x[0]; for the objective function but using GA for example most value end-up becoming close to zero ... any suggestions ? For binaries variables I would try variables from 0.0 to 1.0 and taking 0.5 as threshold for 0 or 1.

Common.h defines GenerateUniformRandomNumber but no code
can be found, although it is called many times ! Any suggestions ?

Hello
May I ask you how do I input a dataset with n variables as model file to this library for run in different kind of techniques?

Thank you

there should be a option to solve permutation problems

Not anything serious, but the link to the tutorial on how to install gcc under OSX in the Installation page from the wiki is broken.

Hello!
Running GA optimization wtih the following parameters:

- target_fn: sphere
- n_variables: 10
- n_agents: 20
- mutation_rate = 0.2
- search_interval: [10, 20] for all variables

yields f(x*) = 0 as optimal value, which is only possible at the space origin, outside the search space. In this case, the expected optimal value should be f(x*) = (10^2)*10 = 1000.

I've created a new stochastic function optimizer which you may be interested in evaluating. It's written in C++, but interfacing with it should not be hard. What I'm interested in is hearing your opinion, based on your benchmarking routine. I'm using a slightly different test function set, and the optimizer solves most problems given at least 50 optimization attempts are performed, up to 2000 iterations each.

Of course, you may include it into your project if you find it interesting.

https://github.com/avaneev/biteopt