Concerns Regarding Initialization of Binary Variables and Support for Custom Initial Population in PyMoo's NSGA2

[haowei2020]

I am currently working on a binary optimization problem where the decision variables are restricted to the values of 0 or 1. I have observed that when initializing the population with the NSGA2 algorithm provided by the PyMoo framework, the variables are initialized as floating-point numbers. I am curious whether this default initialization could potentially impact the optimization process and the quality of the solutions found.

Additionally, I would like to inquire if the PyMoo framework supports the input of a custom initial population where the decision variables are already set as binary values (0 or 1)? If so, could you please provide guidance on how this can be implemented?

[blankjul]

You can write your own sampling method as well and pass it to your algorithm: https://pymoo.org/customization/custom.html

Support for a custom initial population already exists: https://pymoo.org/customization/initialization.html

I hope this helps (please reopen if necessary).