[MNT] isolate scikit-learn

pyproject.toml

@@ -39,7 +39,6 @@ dependencies = [
   "scipy <2.0.0",
   "numpy >=1.18.1, <3.0.0",
   "pandas <3.0.0",
-  "scikit-learn >=0.21, !=0.23.*",
   "tqdm >=4.48.0, <5.0.0",
 ]
 
@@ -53,6 +52,7 @@ test = [
   "pytest == 8.3.5",
   "flake8",
   "pytest-cov",
+  "scikit-learn >=0.21, !=0.23.*",
   "surfaces",
 ]
 

src/gradient_free_optimizers/optimizer_search/bayesian_optimization.py

@@ -6,7 +6,6 @@
 
 from ..search import Search
 from ..optimizers import BayesianOptimizer as _BayesianOptimizer
-from ..optimizers.smb_opt.bayesian_optimization import gaussian_process
 
 
 class BayesianOptimizer(_BayesianOptimizer, Search):
@@ -61,7 +60,7 @@ def __init__(
         max_sample_size: int = 10000000,
         sampling: Dict[Literal["random"], int] = {"random": 1000000},
         replacement: bool = True,
-        gpr=gaussian_process["gp_nonlinear"],
+        gpr=None,
         xi: float = 0.03,
     ):
         super().__init__(

src/gradient_free_optimizers/optimizers/exp_opt/ensemble_optimizer.py

@@ -3,20 +3,12 @@
 # License: MIT License
 
 import numpy as np
-from scipy.stats import norm
 
 from ..smb_opt.smbo import SMBO
 from ..smb_opt.surrogate_models import EnsembleRegressor
 from ..smb_opt.acquisition_function import ExpectedImprovement
 
 
-from sklearn.tree import DecisionTreeRegressor
-from sklearn.ensemble import GradientBoostingRegressor
-from sklearn.svm import SVR
-from sklearn.gaussian_process import GaussianProcessRegressor
-from sklearn.neural_network import MLPRegressor
-
-
 def normalize(array):
     num = array - array.min()
     den = array.max() - array.min()
@@ -41,12 +33,7 @@ def __init__(
         epsilon=0.03,
         distribution="normal",
         n_neighbours=3,
-        estimators=[
-            GradientBoostingRegressor(n_estimators=5),
-            # DecisionTreeRegressor(),
-            # MLPRegressor(),
-            GaussianProcessRegressor(),
-        ],
+        estimators=None,
         xi=0.01,
         warm_start_smbo=None,
         max_sample_size=10000000,
@@ -71,13 +58,27 @@ def __init__(
             replacement=replacement,
         )
         self.estimators = estimators
-        self.regr = EnsembleRegressor(estimators)
         self.xi = xi
         self.warm_start_smbo = warm_start_smbo
         self.max_sample_size = max_sample_size
         self.sampling = sampling
         self.warnings = warnings
 
+        if estimators is None:
+            from sklearn.ensemble import GradientBoostingRegressor
+            from sklearn.gaussian_process import GaussianProcessRegressor
+
+            self._estimators = [
+                GradientBoostingRegressor(n_estimators=5),
+                # DecisionTreeRegressor(),
+                # MLPRegressor(),
+                GaussianProcessRegressor(),
+            ]
+        else:
+            self._estimators = estimators
+
+        self.regr = EnsembleRegressor(self._estimators)
+
         self.init_warm_start_smbo()
 
     def finish_initialization(self):

src/gradient_free_optimizers/optimizers/smb_opt/bayesian_optimization.py

@@ -3,18 +3,13 @@
 # License: MIT License
 
 import numpy as np
-from scipy.stats import norm
 
 
 from .smbo import SMBO
-from .surrogate_models import (
-    GPR_linear,
-    GPR,
-)
 from .acquisition_function import ExpectedImprovement
 
 
-gaussian_process = {"gp_nonlinear": GPR(), "gp_linear": GPR_linear()}
+# gaussian_process = {"gp_nonlinear": GPR(), "gp_linear": GPR_linear()}
 
 
 def normalize(array):
@@ -46,9 +41,9 @@ def __init__(
         nth_process=None,
         warm_start_smbo=None,
         max_sample_size=10000000,
-        sampling={"random": 1000000},
+        sampling=None,
         replacement=True,
-        gpr=gaussian_process["gp_nonlinear"],
+        gpr=None,
         xi=0.03,
     ):
         super().__init__(
@@ -65,7 +60,13 @@ def __init__(
         )
 
         self.gpr = gpr
-        self.regr = gpr
+        if gpr is None:
+            from gradient_free_optimizers.optimizers.smb_opt.surrogate_models import GPR
+            self._gpr = GPR()
+        else:
+            self._gpr = gpr
+
+        self.regr = self._gpr
         self.xi = xi
 
     def finish_initialization(self):

src/gradient_free_optimizers/optimizers/smb_opt/forest_optimizer.py

@@ -53,7 +53,7 @@ def __init__(
         nth_process=None,
         warm_start_smbo=None,
         max_sample_size=10000000,
-        sampling={"random": 1000000},
+        sampling=None,
         replacement=True,
         tree_regressor="extra_tree",
         tree_para={"n_estimators": 100},

src/gradient_free_optimizers/optimizers/smb_opt/smbo.py

@@ -23,7 +23,7 @@ def __init__(
         nth_process=None,
         warm_start_smbo=None,
         max_sample_size=10000000,
-        sampling={"random": 1000000},
+        sampling=None,
         replacement=True,
     ):
         super().__init__(
@@ -40,6 +40,11 @@ def __init__(
         self.sampling = sampling
         self.replacement = replacement
 
+        if sampling is None:
+            self._sampling_dict = {"random": 1000000}
+        else:
+            self._sampling_dict = sampling
+
         self.sampler = InitialSampler(self.conv, max_sample_size)
 
         self.init_warm_start_smbo(warm_start_smbo)
@@ -97,13 +102,13 @@ def wrapper(self, score):
         return wrapper
 
     def _sampling(self, all_pos_comb):
-        if self.sampling is False:
+        if self._sampling_dict is False:
             return all_pos_comb
-        elif "random" in self.sampling:
+        elif "random" in self._sampling_dict:
             return self.random_sampling(all_pos_comb)
 
     def random_sampling(self, pos_comb):
-        n_samples = self.sampling["random"]
+        n_samples = self._sampling_dict["random"]
         n_pos_comb = pos_comb.shape[0]
 
         if n_pos_comb <= n_samples:

src/gradient_free_optimizers/optimizers/smb_opt/tree_structured_parzen_estimators.py

@@ -5,7 +5,6 @@
 
 import numpy as np
 
-from sklearn.neighbors import KernelDensity
 from .smbo import SMBO
 
 
@@ -46,6 +45,8 @@ def __init__(
 
         self.gamma_tpe = gamma_tpe
 
+        from sklearn.neighbors import KernelDensity
+
         kde_para = {
             "kernel": "gaussian",
             "bandwidth": 1,