config updates

tpot2/config/classifiers.py

@@ -6,9 +6,7 @@
 import sklearn
 
 
-def get_LogisticRegression_ConfigurationSpace(n_samples, n_features, random_state):
-
-    dual = n_samples<=n_features
+def get_LogisticRegression_ConfigurationSpace(random_state):
 
     dual = FALSE_SPECIAL_STRING
 
@@ -80,7 +78,7 @@ def get_DecisionTreeClassifier_ConfigurationSpace(n_featues, random_state):
 
     space = {
         'criterion': Categorical("criterion", ['gini', 'entropy']),
-        'max_depth': Integer("max_depth", bounds=(1, 2*n_featues)), #max of 20? log scale?
+        'max_depth': Integer("max_depth", bounds=(1, min(20,2*n_featues))), #max of 20? log scale?
         'min_samples_split': Integer("min_samples_split", bounds=(1, 20)),
         'min_samples_leaf': Integer("min_samples_leaf", bounds=(1, 20)),
         'max_features': Categorical("max_features", [NONE_SPECIAL_STRING, 'sqrt', 'log2']),
@@ -145,10 +143,10 @@ def get_SVC_ConfigurationSpace(random_state):
     return cs
 
 
-def get_RandomForestClassifier_ConfigurationSpace(n_features, random_state):
+def get_RandomForestClassifier_ConfigurationSpace( random_state):
     space = {
             'n_estimators': 128, #as recommended by Oshiro et al. (2012
-            'max_features': Integer("max_features", bounds=(1, max(1, n_features))), #log scale like autosklearn?
+            'max_features': Float("max_features", bounds=(0.01,1)), #log scale like autosklearn?
             'criterion': Categorical("criterion", ['gini', 'entropy']),
             'min_samples_split': Integer("min_samples_split", bounds=(2, 20)),
             'min_samples_leaf': Integer("min_samples_leaf", bounds=(1, 20)),
@@ -211,7 +209,7 @@ def get_ExtraTreesClassifier_ConfigurationSpace(random_state):
     space = {
             'n_estimators': 100,
             'criterion': Categorical("criterion", ["gini", "entropy"]),
-            'max_features': Float("max_features", bounds=(0.05, 1.00)),
+            'max_features': Float("max_features", bounds=(0.01, 1.00)),
             'min_samples_split': Integer("min_samples_split", bounds=(2, 20)),
             'min_samples_leaf': Integer("min_samples_leaf", bounds=(1, 20)),
             'bootstrap': Categorical("bootstrap", [True, False]),
@@ -333,7 +331,7 @@ def get_LinearDiscriminantAnalysis_ConfigurationSpace():
 
 #### Gradient Boosting Classifiers
 
-def get_GradientBoostingClassifier_ConfigurationSpace(n_classes, n_features, random_state):
+def get_GradientBoostingClassifier_ConfigurationSpace(n_classes, random_state):
     early_stop = Categorical("early_stop", ["off", "valid", "train"])
     n_iter_no_change = Integer("n_iter_no_change",bounds=(1,20))
     validation_fraction = Float("validation_fraction", bounds=(0.01, 0.4))
@@ -346,9 +344,9 @@ def get_GradientBoostingClassifier_ConfigurationSpace(n_classes, n_features, ran
         'min_samples_leaf': Integer("min_samples_leaf", bounds=(1, 200)),
         'min_samples_split': Integer("min_samples_split", bounds=(2, 20)),
         'subsample': Float("subsample", bounds=(0.1, 1.0)),
-        'max_features': Integer("max_features", bounds=(1, max(1, n_features))),
+        'max_features': Float("max_features", bounds=(0.01, 1.00)),
         'max_leaf_nodes': Integer("max_leaf_nodes", bounds=(3, 2047)),
-        'max_depth': Integer("max_depth", bounds=(1, 2*n_features)),
+        'max_depth':NONE_SPECIAL_STRING,   # 'max_depth': Integer("max_depth", bounds=(1, 2*n_features)),
         'tol': 1e-4,
     }
 
@@ -400,7 +398,7 @@ def GradientBoostingClassifier_hyperparameter_parser(params):
 
 
 #only difference is l2_regularization
-def get_HistGradientBoostingClassifier_ConfigurationSpace(n_features, random_state):
+def get_HistGradientBoostingClassifier_ConfigurationSpace(random_state):
     early_stop = Categorical("early_stop", ["off", "valid", "train"])
     n_iter_no_change = Integer("n_iter_no_change",bounds=(1,20))
     validation_fraction = Float("validation_fraction", bounds=(0.01, 0.4))
@@ -413,7 +411,7 @@ def get_HistGradientBoostingClassifier_ConfigurationSpace(n_features, random_sta
         'min_samples_leaf': Integer("min_samples_leaf", bounds=(1, 200)),
         'max_features': Float("max_features", bounds=(0.1,1.0)), 
         'max_leaf_nodes': Integer("max_leaf_nodes", bounds=(3, 2047)),
-        'max_depth': Integer("max_depth", bounds=(1, 2*n_features)),
+        'max_depth':NONE_SPECIAL_STRING, # 'max_depth': Integer("max_depth", bounds=(1, 2*n_features)),
         'l2_regularization': Float("l2_regularization", bounds=(1e-10, 1), log=True),
         'tol': 1e-4,
     }

tpot2/config/get_configspace.py

@@ -132,7 +132,7 @@
 
 
 
-def get_configspace(name, n_classes=3, n_samples=100, n_features=100, random_state=None):
+def get_configspace(name, n_classes=3, n_samples=1000, n_features=100, random_state=None):
     match name:
 
         #autoqtl_builtins.py
@@ -158,7 +158,7 @@ def get_configspace(name, n_classes=3, n_samples=100, n_features=100, random_sta
         case "AdaBoostClassifier":
             return classifiers.get_AdaBoostClassifier_ConfigurationSpace(random_state=random_state)
         case "LogisticRegression":
-            return classifiers.get_LogisticRegression_ConfigurationSpace(n_samples=n_samples, n_features=n_features, random_state=random_state)
+            return classifiers.get_LogisticRegression_ConfigurationSpace(random_state=random_state)
         case "KNeighborsClassifier":
             return classifiers.get_KNeighborsClassifier_ConfigurationSpace(n_samples=n_samples)
         case "DecisionTreeClassifier":
@@ -168,11 +168,11 @@ def get_configspace(name, n_classes=3, n_samples=100, n_features=100, random_sta
         case "LinearSVC":
             return classifiers.get_LinearSVC_ConfigurationSpace(random_state=random_state)
         case "RandomForestClassifier":
-            return classifiers.get_RandomForestClassifier_ConfigurationSpace(n_features=n_features, random_state=random_state)
+            return classifiers.get_RandomForestClassifier_ConfigurationSpace(random_state=random_state)
         case "GradientBoostingClassifier":
-            return classifiers.get_GradientBoostingClassifier_ConfigurationSpace(n_classes=n_classes, n_features=n_features, random_state=random_state)
+            return classifiers.get_GradientBoostingClassifier_ConfigurationSpace(n_classes=n_classes, random_state=random_state)
         case "HistGradientBoostingClassifier":
-            return classifiers.get_HistGradientBoostingClassifier_ConfigurationSpace(n_features=n_features, random_state=random_state)
+            return classifiers.get_HistGradientBoostingClassifier_ConfigurationSpace(random_state=random_state)
         case "XGBClassifier":
             return classifiers.get_XGBClassifier_ConfigurationSpace(random_state=random_state)
         case "LGBMClassifier":
@@ -232,7 +232,7 @@ def get_configspace(name, n_classes=3, n_samples=100, n_features=100, random_sta
         case "Perceptron":
             return regressors.get_Perceptron_ConfigurationSpace(random_state=random_state)
         case "DecisionTreeRegressor":
-            return regressors.get_DecisionTreeRegressor_ConfigurationSpace(n_features=n_features, random_state=random_state)
+            return regressors.get_DecisionTreeRegressor_ConfigurationSpace(random_state=random_state)
         case "LinearSVR":
             return regressors.get_LinearSVR_ConfigurationSpace(random_state=random_state)
         case "SVR":
@@ -244,9 +244,9 @@ def get_configspace(name, n_classes=3, n_samples=100, n_features=100, random_sta
         case "ExtraTreesRegressor":
             return regressors.get_ExtraTreesRegressor_ConfigurationSpace(random_state=random_state)
         case "GradientBoostingRegressor":
-            return regressors.get_GradientBoostingRegressor_ConfigurationSpace(n_features=n_features, random_state=random_state)
+            return regressors.get_GradientBoostingRegressor_ConfigurationSpace(random_state=random_state)
         case "HistGradientBoostingRegressor":
-            return regressors.get_HistGradientBoostingRegressor_ConfigurationSpace(n_features=n_features, random_state=random_state)
+            return regressors.get_HistGradientBoostingRegressor_ConfigurationSpace(random_state=random_state)
         case "MLPRegressor":
             return regressors.get_MLPRegressor_ConfigurationSpace(random_state=random_state)
         case "KNeighborsRegressor":

tpot2/config/regressors.py

@@ -219,14 +219,17 @@ def get_Perceptron_ConfigurationSpace(random_state):
 
 
 
-def get_DecisionTreeRegressor_ConfigurationSpace(n_features, random_state):
+def get_DecisionTreeRegressor_ConfigurationSpace(random_state):
     space = {
         'criterion': Categorical("criterion", ['squared_error', 'friedman_mse', 'mae']),
-        'max_depth': Integer("max_depth", bounds=(1, n_features*2)),
+        # 'max_depth': Integer("max_depth", bounds=(1, n_features*2)),
         'min_samples_split': Integer("min_samples_split", bounds=(2, 21)),
         'min_samples_leaf': Integer("min_samples_leaf", bounds=(1, 21)),
     }
 
+    if random_state is not None: #This is required because configspace doesn't allow None as a value
+        space['random_state'] = random_state
+
     return ConfigurationSpace(
         space = space
     )
@@ -382,7 +385,7 @@ def GaussianProcessRegressor_hyperparameter_parser(params):
     return final_params
 
 ###
-def get_GradientBoostingRegressor_ConfigurationSpace(n_features, random_state):
+def get_GradientBoostingRegressor_ConfigurationSpace(random_state):
     early_stop = Categorical("early_stop", ["off", "valid", "train"])
     n_iter_no_change = Integer("n_iter_no_change",bounds=(1,20))
     validation_fraction = Float("validation_fraction", bounds=(0.01, 0.4))
@@ -396,9 +399,9 @@ def get_GradientBoostingRegressor_ConfigurationSpace(n_features, random_state):
         'min_samples_leaf': Integer("min_samples_leaf", bounds=(1, 200)),
         'min_samples_split': Integer("min_samples_split", bounds=(2, 20)),
         'subsample': Float("subsample", bounds=(0.1, 1.0)),
-        'max_features': Integer("max_features", bounds=(1, max(1, n_features))),
+        'max_features': Float("max_features", bounds=(0.01, 1.00)),
         'max_leaf_nodes': Integer("max_leaf_nodes", bounds=(3, 2047)),
-        'max_depth': Integer("max_depth", bounds=(1, 2*n_features)),
+        'max_depth':NONE_SPECIAL_STRING, #'max_depth': Integer("max_depth", bounds=(1, 2*n_features)),
         'tol': 1e-4,
     }
 
@@ -443,7 +446,7 @@ def GradientBoostingRegressor_hyperparameter_parser(params):
     return final_params
 
 #only difference is l2_regularization
-def get_HistGradientBoostingRegressor_ConfigurationSpace(n_features, random_state):
+def get_HistGradientBoostingRegressor_ConfigurationSpace(random_state):
     early_stop = Categorical("early_stop", ["off", "valid", "train"])
     n_iter_no_change = Integer("n_iter_no_change",bounds=(1,20))
     validation_fraction = Float("validation_fraction", bounds=(0.01, 0.4))
@@ -457,7 +460,7 @@ def get_HistGradientBoostingRegressor_ConfigurationSpace(n_features, random_stat
         'min_samples_leaf': Integer("min_samples_leaf", bounds=(1, 200)),
         'max_features': Float("max_features", bounds=(0.1,1.0)), 
         'max_leaf_nodes': Integer("max_leaf_nodes", bounds=(3, 2047)),
-        'max_depth': Integer("max_depth", bounds=(1, 2*n_features)),
+        'max_depth':NONE_SPECIAL_STRING, #'max_depth': Integer("max_depth", bounds=(1, 2*n_features)),
         'l2_regularization': Float("l2_regularization", bounds=(1e-10, 1), log=True),
         'tol': 1e-4,
     }