[FEAT]: Inquiry about Customizing Optimization Starting Point about mealpy HOT 5 CLOSED

Hi @cumttc,

You can pass a matrix of positions with shape (pop_size, n_dims) in a solver like this:
https://mealpy.readthedocs.io/en/latest/pages/general/advance_guide.html#starting-solutions

from mealpy.

Hi@thieupu

Hi @cumttc,

You can pass a matrix of positions with shape (pop_size, n_dims) in a solver like this: https://mealpy.readthedocs.io/en/latest/pages/general/advance_guide.html#starting-solutions

If my initial solution is [1, 2, 3, 4], then the 'starting_solutions' I need to pass in is np.ones((pop_size, 4)) @ np.diag(1, 2, 3, 4).
Is it correct?

from mealpy.

@cumttc ,

Mealpy is population-based algorithms, so the starting_solutions of any algorithm is a matrix of population and problem size.
If your initial solution is [1, 2, 3, 4]. And you have pop_size = 10 (agents).
Then you should set starting_solutions like this:

import numpy as np

pop_size = 10
problem_size = n_dims = 4

initial = np.array([1, 2, 3, 4])
starting = [ initial + np.random.uniform(-0.1, 0.1, n_dims) for _ in range(0, pop_size-1) ]
starting_solutions = np.array( [initial, ] + starting )

You should never set starting_solutions with the same position for all agents. That is not a good practice.

from mealpy.

Thanks for the answer, but one more question:
What should I do if I have additional parameters that are not involved in the optimization, but need to be passed into the objective function?

from mealpy.

@cumttc ,

Then you should define a custom problem like this:
https://mealpy.readthedocs.io/en/latest/pages/general/advance_guide.html#custom-problem

You can pass anything you want into that class.

from mealpy.