Convergence on local optima

[Nistha-gif]

Dear Radoslaw,
I am using Deap framwork for solving a variant of multi-depot VRP using PSO. I have observed that the algorithm converges on local optima. As you have solved the problem using GA and PSO. I would really appreciate it if you could update the code of PSO to reduce or eliminate the possibility of convergence on local optima.

[radoslawik]

Hello,

Due to a lack of time, I think I will not be able to make any improvements to this project soon. You could try to play with the values of parameters and test them against your specific problem. If you are still not happy with the result, you could try to add new random particles to every iteration (and remove the same amount of particles from the current population if you want to preserve the same size) or add more randomness when calculating their move.

The balance between exploration and exploitation is not a trivial thing, I really like the sentence on the wikipedia:

...determining convergence capabilities of different PSO algorithms and parameters still depends on empirical results.

Good luck with your project! If you find a way to improve the algorithm you can share your thoughts here (or create a pull request if you are motivated)