gradient_free_optimizers.optimizers.local.repulsing_hill_climbing_optimizer

Complexity

Total Complexity

4

Size/Duplication

Total Lines	39
Duplicated Lines	0 %

Importance

Changes

0

3 Methods

Rating	Name	Duplication	Size	Complexity
A	RepulsingHillClimbingOptimizer.evaluate()	0	7	2
A	RepulsingHillClimbingOptimizer.__init__()	0	15	1
A	RepulsingHillClimbingOptimizer.iterate()	0	4	1

# Author: Simon Blanke
# Email: simon.blanke@yahoo.com
# License: MIT License


from . import HillClimbingOptimizer
from ...search import Search


class RepulsingHillClimbingOptimizer(HillClimbingOptimizer, Search):
    def __init__(
        self,
        search_space,
        epsilon=0.03,
        distribution="normal",
        n_neighbours=3,
        repulsion_factor=5,
        rand_rest_p=0.03,
    ):
        super().__init__(search_space, epsilon, distribution, n_neighbours)

        self.tabus = []
        self.repulsion_factor = repulsion_factor
        self.rand_rest_p = rand_rest_p
        self.epsilon_mod = 1

    @HillClimbingOptimizer.track_nth_iter
    @HillClimbingOptimizer.random_restart
    def iterate(self):
        return self._move_climb(self.pos_current, epsilon_mod=self.epsilon_mod)

    def evaluate(self, score_new):
        super().evaluate(score_new)

        if score_new <= self.score_current:
            self.epsilon_mod = self.repulsion_factor
        else:
            self.epsilon_mod = 1