Merge branch 'dev' of https://github.com/SimonBlanke/Hyperactive into dev

Author: SimonBlanke

examples/meta_opt_extensive_example.py

@@ -0,0 +1,86 @@
+import numpy as np
+import random
+from sklearn.model_selection import cross_val_score
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import load_breast_cancer
+from hyperactive import Hyperactive
+
+data = load_breast_cancer()
+X, y = data.data, data.target
+
+X_list, y_list = [X], [y]
+
+def data_aug(X, y, sample_multi=5, feature_multi=5):
+    X_list, y_list = [], []
+
+    n_samples = X.shape[0]
+    n_features = X.shape[1]
+
+    for sample in range(1, sample_multi+1):
+        idx_sample = np.random.randint(n_samples, size=int(n_samples / sample))
+
+        for feature in range(1, feature_multi+1):
+            idx_feature = np.random.randint(n_features, size=int(n_features / feature))
+
+            X_temp_ = X[idx_sample, :]
+            X_temp = X_temp_[:, idx_feature]
+            y_temp = y[idx_sample]
+
+            X_list.append(X_temp)
+            y_list.append(y_temp)
+
+    return X_list, y_list
+
+def model(para, X, y):
+    model = DecisionTreeClassifier(
+        max_depth=para["max_depth"],
+        min_samples_split=para["min_samples_split"],
+        min_samples_leaf=para["min_samples_leaf"],
+    )
+    scores = cross_val_score(model, X, y, cv=3)
+
+    return scores.mean()
+
+search_config_model = {
+    model: {
+        "max_depth": range(2, 50),
+        "min_samples_split": range(2, 50),
+        "min_samples_leaf": range(1, 50),
+    }
+}
+
+
+def meta_opt(para, X_list, y_list):
+    scores = []
+    
+    for X, y in zip(X_list, y_list):
+        X_list, y_list = data_aug(X, y, sample_multi=3, feature_multi=3)
+
+        for X, y in zip(X_list, y_list):
+
+            for n_iter in [10, 25, 50, 100]:
+                opt = Hyperactive(
+                    search_config_model,
+                    optimizer={
+                        "ParticleSwarm": {"inertia": para["inertia"], "cognitive_weight": para["cognitive_weight"], "social_weight": para["social_weight"]}
+                    }, 
+                    n_iter=n_iter,
+                    verbosity=None,
+                )
+                opt.search(X, y)
+                score = opt.score_best
+                scores.append(score)
+
+    return np.array(scores).mean()
+
+
+search_config_meta = {
+    meta_opt: {
+        "inertia": np.arange(0, 1, 0.01),
+        "cognitive_weight": np.arange(0, 1, 0.01),
+        "social_weight": np.arange(0, 1, 0.01),
+    }
+}
+
+opt = Hyperactive(search_config_meta, optimizer="Bayesian", n_iter=30)
+opt.search(X_list, y_list)