SimonBlanke /
Gradient-Free-Optimizers
| Total Complexity | 6 |
| Total Lines | 80 |
| Duplicated Lines | 56.25 % |
| Changes | 0 | ||
Duplicate code is one of the most pungent code smells. A rule that is often used is to re-structure code once it is duplicated in three or more places.
Common duplication problems, and corresponding solutions are:
| 1 | # Author: Simon Blanke |
||
| 2 | # Email: [email protected] |
||
| 3 | # License: MIT License |
||
| 4 | |||
| 5 | import numpy as np |
||
| 6 | from scipy.stats import norm |
||
| 7 | |||
| 8 | |||
| 9 | from .smbo import SMBO |
||
| 10 | from .surrogate_models import ( |
||
| 11 | RandomForestRegressor, |
||
| 12 | ExtraTreesRegressor, |
||
| 13 | GradientBoostingRegressor, |
||
| 14 | ) |
||
| 15 | from .acquisition_function import ExpectedImprovement |
||
| 16 | |||
| 17 | |||
| 18 | tree_regressor_dict = { |
||
| 19 | "random_forest": RandomForestRegressor, |
||
| 20 | "extra_tree": ExtraTreesRegressor, |
||
| 21 | "gradient_boost": GradientBoostingRegressor, |
||
| 22 | } |
||
| 23 | |||
| 24 | |||
| 25 | def normalize(array): |
||
| 26 | num = array - array.min() |
||
| 27 | den = array.max() - array.min() |
||
| 28 | |||
| 29 | if den == 0: |
||
| 30 | return np.random.random_sample(array.shape) |
||
| 31 | else: |
||
| 32 | return ((num / den) + 0) / 1 |
||
| 33 | |||
| 34 | |||
| 35 | View Code Duplication | class ForestOptimizer(SMBO): |
|
|
Duplication
introduced
by
|
|||
| 36 | name = "Forest Optimization" |
||
| 37 | _name_ = "forest_optimization" |
||
| 38 | """Based on the forest-optimizer in the scikit-optimize package""" |
||
| 39 | |||
| 40 | def __init__( |
||
| 41 | self, |
||
| 42 | *args, |
||
| 43 | tree_regressor="extra_tree", |
||
| 44 | tree_para={"n_estimators": 100}, |
||
| 45 | xi=0.03, |
||
| 46 | warm_start_smbo=None, |
||
| 47 | max_sample_size=10000000, |
||
| 48 | sampling={"random": 1000000}, |
||
| 49 | warnings=100000000, |
||
| 50 | **kwargs |
||
| 51 | ): |
||
| 52 | super().__init__(*args, **kwargs) |
||
| 53 | self.tree_regressor = tree_regressor |
||
| 54 | self.tree_para = tree_para |
||
| 55 | self.regr = tree_regressor_dict[tree_regressor](**self.tree_para) |
||
| 56 | self.xi = xi |
||
| 57 | self.warm_start_smbo = warm_start_smbo |
||
| 58 | self.max_sample_size = max_sample_size |
||
| 59 | self.sampling = sampling |
||
| 60 | self.warnings = warnings |
||
| 61 | |||
| 62 | self.init_warm_start_smbo() |
||
| 63 | |||
| 64 | def _expected_improvement(self): |
||
| 65 | all_pos_comb = self._all_possible_pos() |
||
| 66 | self.pos_comb = self._sampling(all_pos_comb) |
||
| 67 | |||
| 68 | acqu_func = ExpectedImprovement(self.regr, self.pos_comb, self.xi) |
||
| 69 | return acqu_func.calculate(self.X_sample, self.Y_sample) |
||
| 70 | |||
| 71 | def _training(self): |
||
| 72 | X_sample = np.array(self.X_sample) |
||
| 73 | Y_sample = np.array(self.Y_sample) |
||
| 74 | |||
| 75 | if len(Y_sample) == 0: |
||
| 76 | return self.move_random() |
||
| 77 | |||
| 78 | Y_sample = normalize(Y_sample).reshape(-1, 1) |
||
| 79 | self.regr.fit(X_sample, Y_sample) |
||
| 80 |
Loading history...