How to understand the way of Generating weight vectors about platypus HOT 3 CLOSED

Hi and welcome!

Yes, you're correct that they didn't use random_weights in the original MOEA/D.

In Li, H. and Zhang, Q. "Multiobjective Optimization problems with Complicated Pareto Sets, MOEA/D and NSGA-II." IEEE Transactions on Evolutionary Computation, 13(2):284-302, 2009, their approach is equivalent to using normal_boundary_weights with divisions_outer = H and divisions_inner = 0.

In Zhang, Q., et al. "The Performance of a New Version of MOEA/D on CEC09 Unconstrained MOP Test Instances." IEEE Congress on Evolutionary Computation, 2009, they introduced MOEA/D-DRA which supports dynamic resource allocation (the update_utility argument is used for this). In this paper, they use the random_weights method.

So why would we use random_weights? Well, the number of weights produced by normal_boundary_weights grows very quickly as the number of objectives increases.

>>> len(platypus.normal_boundary_weights(2, divisions_outer=100))
101
>>> len(platypus.normal_boundary_weights(3, divisions_outer=100))
5151
>>> len(platypus.normal_boundary_weights(4, divisions_outer=100))
176851

Which means we need to adjust the value of H for every objective count. Using random_weights is much easier...just tell it how many weights to generate.

Please note that while writing this response, I found a bug in how weight_generator was being handled and committed a fix. With this fix, you would use the different weight generators as follows:

MOEAD(problem, population_size=100, weight_generator=random_weights)

or

MOEAD(problem, divisions_outer=24, weight_generator=normal_boundary_weights)

from platypus.

Excellent !! I really appreciate your help !!!
Now I must gain a better understanding of both the generating methods and the code itself.

BTW, Do you prefer my creating issue or sending emails if I encountered other uncertainty about the code in the future ?
I know it may not be proper to create such an issue.

from platypus.

No problem! I prefer having the issues created here since others may have the same or similar questions. Closing this issue.

from platypus.