|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
from collections.abc import Iterable, Callable |
|
6
|
|
|
from typing import Union, Dict, Type |
|
7
|
|
|
|
|
8
|
|
|
from sklearn.base import BaseEstimator, clone |
|
9
|
|
|
from sklearn.metrics import check_scoring |
|
10
|
|
|
from sklearn.utils.validation import indexable, _check_method_params |
|
11
|
|
|
|
|
12
|
|
|
from sklearn.base import BaseEstimator as SklearnBaseEstimator |
|
13
|
|
|
from sklearn.model_selection import BaseCrossValidator |
|
14
|
|
|
|
|
15
|
|
|
from hyperactive import Hyperactive |
|
16
|
|
|
|
|
17
|
|
|
from .objective_function_adapter import ObjectiveFunctionAdapter |
|
18
|
|
|
from .best_estimator import BestEstimator as _BestEstimator_ |
|
19
|
|
|
from .checks import Checks |
|
20
|
|
|
from ...optimizers import RandomSearchOptimizer |
|
21
|
|
|
|
|
22
|
|
|
|
|
23
|
|
|
class HyperactiveSearchCV(BaseEstimator, _BestEstimator_, Checks): |
|
24
|
|
|
""" |
|
25
|
|
|
HyperactiveSearchCV class for hyperparameter tuning using cross-validation with sklearn estimators. |
|
26
|
|
|
|
|
27
|
|
|
Parameters: |
|
28
|
|
|
- estimator: SklearnBaseEstimator |
|
29
|
|
|
The estimator to be tuned. |
|
30
|
|
|
- params_config: Dict[str, list] |
|
31
|
|
|
Dictionary containing the hyperparameter search space. |
|
32
|
|
|
- optimizer: Union[str, Type[RandomSearchOptimizer]], optional |
|
33
|
|
|
The optimizer to be used for hyperparameter search, default is "default". |
|
34
|
|
|
- n_iter: int, optional |
|
35
|
|
|
Number of parameter settings that are sampled, default is 100. |
|
36
|
|
|
- scoring: Callable | str | None, optional |
|
37
|
|
|
Scoring method to evaluate the predictions on the test set. |
|
38
|
|
|
- n_jobs: int, optional |
|
39
|
|
|
Number of jobs to run in parallel, default is 1. |
|
40
|
|
|
- random_state: int | None, optional |
|
41
|
|
|
Random seed for reproducibility. |
|
42
|
|
|
- refit: bool, optional |
|
43
|
|
|
Refit the best estimator with the entire dataset, default is True. |
|
44
|
|
|
- cv: int | "BaseCrossValidator" | Iterable | None, optional |
|
45
|
|
|
Determines the cross-validation splitting strategy. |
|
46
|
|
|
|
|
47
|
|
|
Methods: |
|
48
|
|
|
- fit(X, y, **fit_params) |
|
49
|
|
|
Fit the estimator and tune hyperparameters. |
|
50
|
|
|
- score(X, y, **params) |
|
51
|
|
|
Return the score of the best estimator on the input data. |
|
52
|
|
|
""" |
|
53
|
|
|
|
|
54
|
|
|
_required_parameters = ["estimator", "optimizer", "params_config"] |
|
55
|
|
|
|
|
56
|
|
|
def __init__( |
|
57
|
|
|
self, |
|
58
|
|
|
estimator: "SklearnBaseEstimator", |
|
59
|
|
|
params_config: Dict[str, list], |
|
60
|
|
|
optimizer: Union[str, Type[RandomSearchOptimizer]] = "default", |
|
61
|
|
|
n_iter: int = 100, |
|
62
|
|
|
*, |
|
63
|
|
|
scoring: Union[Callable, str, None] = None, |
|
64
|
|
|
n_jobs: int = 1, |
|
65
|
|
|
random_state: Union[int, None] = None, |
|
66
|
|
|
refit: bool = True, |
|
67
|
|
|
cv=None, |
|
68
|
|
|
): |
|
69
|
|
|
super().__init__() |
|
70
|
|
|
|
|
71
|
|
|
self.estimator = estimator |
|
72
|
|
|
self.params_config = params_config |
|
73
|
|
|
self.optimizer = optimizer |
|
74
|
|
|
self.n_iter = n_iter |
|
75
|
|
|
self.scoring = scoring |
|
76
|
|
|
self.n_jobs = n_jobs |
|
77
|
|
|
self.random_state = random_state |
|
78
|
|
|
self.refit = refit |
|
79
|
|
|
self.cv = cv |
|
80
|
|
|
|
|
81
|
|
|
def _refit(self, X, y=None, **fit_params): |
|
82
|
|
|
self.best_estimator_ = clone(self.estimator).set_params( |
|
83
|
|
|
**clone(self.best_params_, safe=False) |
|
84
|
|
|
) |
|
85
|
|
|
|
|
86
|
|
|
self.best_estimator_.fit(X, y, **fit_params) |
|
87
|
|
|
return self |
|
88
|
|
|
|
|
89
|
|
|
def _check_data(self, X, y): |
|
90
|
|
|
X, y = indexable(X, y) |
|
91
|
|
|
if hasattr(self, "_validate_data"): |
|
92
|
|
|
validate_data = self._validate_data |
|
93
|
|
|
else: |
|
94
|
|
|
from sklearn.utils.validation import validate_data |
|
95
|
|
|
|
|
96
|
|
|
return validate_data(X, y) |
|
97
|
|
|
|
|
98
|
|
|
@Checks.verify_fit |
|
99
|
|
|
def fit(self, X, y, **fit_params): |
|
100
|
|
|
""" |
|
101
|
|
|
Fit the estimator using the provided training data. |
|
102
|
|
|
|
|
103
|
|
|
Parameters: |
|
104
|
|
|
- X: array-like or sparse matrix, shape (n_samples, n_features) |
|
105
|
|
|
The training input samples. |
|
106
|
|
|
- y: array-like, shape (n_samples,) or (n_samples, n_outputs) |
|
107
|
|
|
The target values. |
|
108
|
|
|
- **fit_params: dict of string -> object |
|
109
|
|
|
Additional fit parameters. |
|
110
|
|
|
|
|
111
|
|
|
Returns: |
|
112
|
|
|
- self: object |
|
113
|
|
|
Returns the instance itself. |
|
114
|
|
|
""" |
|
115
|
|
|
|
|
116
|
|
|
X, y = self._check_data(X, y) |
|
117
|
|
|
|
|
118
|
|
|
fit_params = _check_method_params(X, params=fit_params) |
|
119
|
|
|
self.scorer_ = check_scoring(self.estimator, scoring=self.scoring) |
|
120
|
|
|
|
|
121
|
|
|
objective_function_adapter = ObjectiveFunctionAdapter( |
|
122
|
|
|
self.estimator, |
|
123
|
|
|
) |
|
124
|
|
|
objective_function_adapter.add_dataset(X, y) |
|
125
|
|
|
objective_function_adapter.add_validation(self.scorer_, self.cv) |
|
126
|
|
|
objective_function = objective_function_adapter.objective_function |
|
127
|
|
|
|
|
128
|
|
|
hyper = Hyperactive(verbosity=False) |
|
129
|
|
|
hyper.add_search( |
|
130
|
|
|
objective_function, |
|
131
|
|
|
search_space=self.params_config, |
|
132
|
|
|
optimizer=self.optimizer, |
|
133
|
|
|
n_iter=self.n_iter, |
|
134
|
|
|
n_jobs=self.n_jobs, |
|
135
|
|
|
random_state=self.random_state, |
|
136
|
|
|
) |
|
137
|
|
|
hyper.run() |
|
138
|
|
|
|
|
139
|
|
|
self.best_params_ = hyper.best_para(objective_function) |
|
140
|
|
|
self.best_score_ = hyper.best_score(objective_function) |
|
141
|
|
|
self.search_data_ = hyper.search_data(objective_function) |
|
142
|
|
|
|
|
143
|
|
|
if self.refit: |
|
144
|
|
|
self._refit(X, y, **fit_params) |
|
145
|
|
|
|
|
146
|
|
|
return self |
|
147
|
|
|
|
|
148
|
|
|
def score(self, X, y=None, **params): |
|
149
|
|
|
""" |
|
150
|
|
|
Calculate the score of the best estimator on the input data. |
|
151
|
|
|
|
|
152
|
|
|
Parameters: |
|
153
|
|
|
- X: array-like or sparse matrix of shape (n_samples, n_features) |
|
154
|
|
|
The input samples. |
|
155
|
|
|
- y: array-like of shape (n_samples,), default=None |
|
156
|
|
|
The target values. |
|
157
|
|
|
- **params: dict |
|
158
|
|
|
Additional parameters to be passed to the scoring function. |
|
159
|
|
|
|
|
160
|
|
|
Returns: |
|
161
|
|
|
- float |
|
162
|
|
|
The score of the best estimator on the input data. |
|
163
|
|
|
""" |
|
164
|
|
|
|
|
165
|
|
|
return self.scorer_(self.best_estimator_, X, y, **params) |
|
166
|
|
|
|
|
167
|
|
|
@property |
|
168
|
|
|
def fit_successful(self): |
|
169
|
|
|
self._fit_successful |
|
170
|
|
|
|
SimonBlanke /
Hyperactive
