for testing and deploying your application
for finding and fixing issues
for empowering human code reviews
# Author: Simon Blanke
# Email: [email protected]
# License: MIT License
import random
import numpy as np
from . import HillClimbingOptimizer
from ...search import Search
from scipy.spatial.distance import euclidean
def gaussian(distance, sig, sigma_factor=1):
return (
sigma_factor
* sig
* np.exp(-np.power(distance, 2.0) / (sigma_factor * np.power(sig, 2.0)))
)
class TabuOptimizer(HillClimbingOptimizer, Search):
def __init__(self, search_space, tabu_factor=3):
super().__init__(search_space)
self.tabus = []
self.tabu_factor = tabu_factor
self.epsilon_mod = 1
def iterate(self):
return self._move_climb(self.pos_current)
def evaluate(self, score_new):
super().evaluate(score_new)
if score_new <= self.score_current:
self.epsilon_mod = self.tabu_factor
else:
SimonBlanke /
Gradient-Free-Optimizers
