|
1
|
|
|
"""Grid optimizer.""" |
|
2
|
|
|
# copyright: hyperactive developers, MIT License (see LICENSE file) |
|
3
|
|
|
|
|
4
|
|
|
from .._adapters._base_optuna_adapter import _BaseOptunaAdapter |
|
5
|
|
|
|
|
6
|
|
|
|
|
7
|
|
|
class GridOptimizer(_BaseOptunaAdapter): |
|
8
|
|
|
"""Grid search optimizer. |
|
9
|
|
|
|
|
10
|
|
|
Parameters |
|
11
|
|
|
---------- |
|
12
|
|
|
param_space : dict[str, tuple or list or optuna distributions] |
|
13
|
|
|
The search space to explore. Dictionary with parameter names |
|
14
|
|
|
as keys and either tuples/lists of (low, high) or |
|
15
|
|
|
optuna distribution objects as values. |
|
16
|
|
|
n_trials : int, default=100 |
|
17
|
|
|
Number of optimization trials. |
|
18
|
|
|
initialize : dict[str, int], default=None |
|
19
|
|
|
The method to generate initial positions. A dictionary with |
|
20
|
|
|
the following key literals and the corresponding value type: |
|
21
|
|
|
{"grid": int, "vertices": int, "random": int, "warm_start": list[dict]} |
|
22
|
|
|
random_state : None, int, default=None |
|
23
|
|
|
If None, create a new random state. If int, create a new random state |
|
24
|
|
|
seeded with the value. |
|
25
|
|
|
early_stopping : int, default=None |
|
26
|
|
|
Number of trials after which to stop if no improvement. |
|
27
|
|
|
max_score : float, default=None |
|
28
|
|
|
Maximum score threshold. Stop optimization when reached. |
|
29
|
|
|
search_space : dict, default=None |
|
30
|
|
|
Explicit search space for grid search. |
|
31
|
|
|
experiment : BaseExperiment, optional |
|
32
|
|
|
The experiment to optimize parameters for. |
|
33
|
|
|
Optional, can be passed later via ``set_params``. |
|
34
|
|
|
|
|
35
|
|
|
Examples |
|
36
|
|
|
-------- |
|
37
|
|
|
Basic usage of GridOptimizer with a scikit-learn experiment: |
|
38
|
|
|
|
|
39
|
|
|
>>> from hyperactive.experiment.integrations import SklearnCvExperiment |
|
40
|
|
|
>>> from hyperactive.opt.optuna import GridOptimizer |
|
41
|
|
|
>>> from sklearn.datasets import load_iris |
|
42
|
|
|
>>> from sklearn.svm import SVC |
|
43
|
|
|
>>> X, y = load_iris(return_X_y=True) |
|
44
|
|
|
>>> sklearn_exp = SklearnCvExperiment(estimator=SVC(), X=X, y=y) |
|
45
|
|
|
>>> param_space = { |
|
46
|
|
|
... "C": [0.01, 0.1, 1, 10], |
|
47
|
|
|
... "gamma": [0.0001, 0.01, 0.1, 1], |
|
48
|
|
|
... } |
|
49
|
|
|
>>> optimizer = GridOptimizer( |
|
50
|
|
|
... param_space=param_space, n_trials=50, experiment=sklearn_exp |
|
51
|
|
|
... ) |
|
52
|
|
|
>>> best_params = optimizer.run() |
|
53
|
|
|
""" |
|
54
|
|
|
|
|
55
|
|
|
_tags = { |
|
56
|
|
|
"info:name": "Grid Optimizer", |
|
57
|
|
|
"info:local_vs_global": "global", |
|
58
|
|
|
"info:explore_vs_exploit": "explore", |
|
59
|
|
|
"info:compute": "low", |
|
60
|
|
|
"python_dependencies": ["optuna"], |
|
61
|
|
|
} |
|
62
|
|
|
|
|
63
|
|
|
def __init__( |
|
64
|
|
|
self, |
|
65
|
|
|
param_space=None, |
|
66
|
|
|
n_trials=100, |
|
67
|
|
|
initialize=None, |
|
68
|
|
|
random_state=None, |
|
69
|
|
|
early_stopping=None, |
|
70
|
|
|
max_score=None, |
|
71
|
|
|
search_space=None, |
|
72
|
|
|
experiment=None, |
|
73
|
|
|
): |
|
74
|
|
|
self.search_space = search_space |
|
75
|
|
|
|
|
76
|
|
|
super().__init__( |
|
77
|
|
|
param_space=param_space, |
|
78
|
|
|
n_trials=n_trials, |
|
79
|
|
|
initialize=initialize, |
|
80
|
|
|
random_state=random_state, |
|
81
|
|
|
early_stopping=early_stopping, |
|
82
|
|
|
max_score=max_score, |
|
83
|
|
|
experiment=experiment, |
|
84
|
|
|
) |
|
85
|
|
|
|
|
86
|
|
|
def _get_optimizer(self): |
|
87
|
|
|
"""Get the Grid optimizer. |
|
88
|
|
|
|
|
89
|
|
|
Returns |
|
90
|
|
|
------- |
|
91
|
|
|
optimizer |
|
92
|
|
|
The Optuna GridOptimizer instance |
|
93
|
|
|
""" |
|
94
|
|
|
import optuna |
|
95
|
|
|
|
|
96
|
|
|
# Convert param_space to Optuna search space format if needed |
|
97
|
|
|
search_space = self.search_space |
|
98
|
|
|
if search_space is None and self.param_space is not None: |
|
99
|
|
|
search_space = {} |
|
100
|
|
|
for key, space in self.param_space.items(): |
|
101
|
|
|
if isinstance(space, list): |
|
102
|
|
|
search_space[key] = space |
|
103
|
|
|
elif isinstance(space, (tuple,)) and len(space) == 2: |
|
104
|
|
|
# Convert range to discrete list for grid search |
|
105
|
|
|
low, high = space |
|
106
|
|
|
if isinstance(low, int) and isinstance(high, int): |
|
107
|
|
|
search_space[key] = list(range(low, high + 1)) |
|
108
|
|
|
else: |
|
109
|
|
|
# Create a reasonable grid for continuous spaces |
|
110
|
|
|
import numpy as np |
|
111
|
|
|
|
|
112
|
|
|
search_space[key] = np.linspace(low, high, 10).tolist() |
|
113
|
|
|
|
|
114
|
|
|
return optuna.samplers.GridSampler(search_space) |
|
115
|
|
|
|
|
116
|
|
|
@classmethod |
|
117
|
|
|
def get_test_params(cls, parameter_set="default"): |
|
118
|
|
|
"""Return testing parameter settings for the optimizer.""" |
|
119
|
|
|
from sklearn.datasets import load_iris |
|
120
|
|
|
from sklearn.neighbors import KNeighborsClassifier |
|
121
|
|
|
from sklearn.svm import SVC |
|
122
|
|
|
|
|
123
|
|
|
from hyperactive.experiment.integrations import SklearnCvExperiment |
|
124
|
|
|
|
|
125
|
|
|
X, y = load_iris(return_X_y=True) |
|
126
|
|
|
|
|
127
|
|
|
# Test case 1: Basic continuous parameters (converted to discrete) |
|
128
|
|
|
svm_exp = SklearnCvExperiment(estimator=SVC(), X=X, y=y) |
|
129
|
|
|
param_space_1 = { |
|
130
|
|
|
"C": [0.01, 0.1, 1, 10], |
|
131
|
|
|
"gamma": [0.0001, 0.01, 0.1, 1], |
|
132
|
|
|
} |
|
133
|
|
|
|
|
134
|
|
|
# Test case 2: Mixed categorical and discrete parameters |
|
135
|
|
|
knn_exp = SklearnCvExperiment(estimator=KNeighborsClassifier(), X=X, y=y) |
|
136
|
|
|
param_space_2 = { |
|
137
|
|
|
"n_neighbors": [1, 3, 5, 7], # Discrete integers |
|
138
|
|
|
"weights": ["uniform", "distance"], # Categorical |
|
139
|
|
|
"metric": ["euclidean", "manhattan"], # Categorical |
|
140
|
|
|
"p": [1, 2], # Discrete for minkowski |
|
141
|
|
|
} |
|
142
|
|
|
|
|
143
|
|
|
# Test case 3: Small exhaustive grid (tests complete enumeration) |
|
144
|
|
|
param_space_3 = { |
|
145
|
|
|
"C": [0.1, 1], # 2 values |
|
146
|
|
|
"kernel": ["rbf", "linear"], # 2 values |
|
147
|
|
|
} |
|
148
|
|
|
# Total: 2 x 2 = 4 combinations, n_trials should cover all |
|
149
|
|
|
|
|
150
|
|
|
return [ |
|
151
|
|
|
{ |
|
152
|
|
|
"param_space": param_space_1, |
|
153
|
|
|
"n_trials": 10, |
|
154
|
|
|
"experiment": svm_exp, |
|
155
|
|
|
}, |
|
156
|
|
|
{ |
|
157
|
|
|
"param_space": param_space_2, |
|
158
|
|
|
"n_trials": 15, |
|
159
|
|
|
"experiment": knn_exp, |
|
160
|
|
|
}, |
|
161
|
|
|
{ |
|
162
|
|
|
"param_space": param_space_3, |
|
163
|
|
|
"n_trials": 4, # Exact number for exhaustive search |
|
164
|
|
|
"experiment": svm_exp, |
|
165
|
|
|
}, |
|
166
|
|
|
] |
|
167
|
|
|
|
SimonBlanke /
Hyperactive
