Exporting Best Fitness Variables from Surrogate Algorithm

Hi all,

I’m trying to export only the best fitness variables (5 variables) from a surrogate optimization algorithm. Currently, I’m able to export the variables at each generation, but I would like to export only those that correspond to an improvement in the best fitness value.

Since the best value doesn’t necessarily change at every generation, I’d prefer to save only the variables when a new best is found.

Does anyone know how to implement this or have a suggestion for tracking and exporting only the best-performing variables throughout the optimization?

Best regards,

Mirhan

Answer 1

[Matt J]

The OutputFcn option gives you access to incumbentX,

https://www.mathworks.com/help/gads/surrogate-optimization-options.html#mw_26e25441-83d3-4b43-a67a-089e5cc92634

[Mirhan]

Dear Matt,

Thank you for your response. I must have overlooked that—after your message, I checked again, and it seems to give exactly what I need.

That said, I still had to remove the redundant rows using:

x_history_unique = unique(x_history, 'rows');.

Best regards,

Mirhan