Program crashes on archi.wait_check();

[nconsulting]

Hi,

I have written in C++ a non linear integer optimization problem with 1 objective function and 3 inequality constraints. I have used the algorithms sga and nsga2. Both crash when archi.wait_check(); is called. The crash happened when I added the 3 inequality constraints. It creates when population fine, but when it has to evolve and wait_check, it crashes.

I just tested with the ihs algorithm and using that algorithm my program doesn't crash. However, it is not usable because after the evolve step the decision variables become doubles instead of just integers, which is not usable for me.

Any ideas where I should start debugging my problem and what this issue may cause?

Kind regards,

Nick

[bluescarni]

Can you give more details? How does it crash?

archi.wait_check() will re-raise any exception that was thrown during the asynchronous evolution, so perhaps the algorithms sga/nsga2 are raising some error. Did you check the documentation of those algorithms?

https://esa.github.io/pagmo2/docs/algorithm_list.html

[nconsulting]

Thank you for your reply. I added a screenshot of the error that I get (it is not very descriptive).

I just checked the documentation.

sga - evolve() throws an error when the problem is constrained nsga2 - evolve() throws an error when the problem is single objective

Since my problem has only one objective evolve throws an error and archi.wait_check() makes my program crash. And since my problem has constraints sga isn't working and makes my program crash. Are the above 2 statements correct?

So that only leaves ihs - for a single objective, constrained integer problem . Correct?

Is there a way to make sure that ihs only selects integers for the decision variables? Because now when I start the program with the ihs algorithm the decision variables are all integer values, but after an evolve step they become (mostly) doubles.

[bluescarni]

Ah yeah I see, error messages on uncaught exceptions are usually more descriptive in linux :)

I will have to ping again @darioizzo regarding your questions, as he is the algorithmic expert.

On a side note, we usually encourage users to join the chat channel rather than opening new issues for questions on pagmo/pygmo's usage:

https://gitter.im/pagmo2/Lobby

[darioizzo]

The integer variables have to be defined by implementing in the UDP the get_nix method, please see the documentation (https://esa.github.io/pagmo2/docs/python/py_problem.html) and tutorials (https://esa.github.io/pagmo2/docs/python/tutorials/coding_udp_minlp.html)

[bluescarni]

I'll go ahead and close this, as this is not a pagmo crash.