+ compared implementations with sklearn

! missing parameter comparaison

src/sampling.py

@@ -0,0 +1,113 @@
+from naive_bayes import *
+from sklearn.model_selection import train_test_split
+from sklearn.linear_model import LogisticRegression
+from sklearn.naive_bayes import GaussianNB
+from metrics import compute_metrics
+import softmax
+from matplotlib import pyplot as plt
+from main import FEAT, LABELS
+import pandas as pd
+import numpy as np
+
+# ---------------------------------------------------------------------------
+# ECHANTILLONNAGE
+
+# Prendre 50 échantillons de chaque classe
+SAMPLES = 50
+parameters = get_distrib_parameters(FEAT, LABELS)
+classes = parameters.keys()
+
+sampled_data = []
+sampled_labels = []
+
+# On génère des échantillons pour chaque classe
+for y in classes:
+    class_samples = []
+
+    for variable_params in parameters[y]:
+
+        mean, std = variable_params
+        samples = np.random.normal(mean, std, SAMPLES)
+        class_samples.append(samples)
+
+    class_samples = np.column_stack(class_samples)
+    sampled_data.append(class_samples)
+    sampled_labels.extend([y]*SAMPLES)
+
+# On concatène les échantillons
+sampled_data = np.vstack(sampled_data)
+sampled_labels = np.array(sampled_labels)
+# ---------------------------------------------------------------------------
+
+
+# ---------------------------------------------------------------------------
+# COMPARAISON
+
+for c in classes:
+    means, stds = zip(*parameters[c])
+    print(f"Classe {c} réelle")
+    print(f"Mean: {means}")
+    print(f"Std: {stds}")
+
+    mean_sampled = np.mean(sampled_data[sampled_labels == c], axis=0)
+    std_sampled = np.std(sampled_data[sampled_labels == c], axis=0)
+
+    print(f"Classe {c} échantillonnée")
+    print(f"Mean: {mean_sampled}")
+    print(f"Std: {std_sampled}")
+
+    vars = [0, 1, 2, 3]    
+
+
+print("\n\n")
+
+X_train, X_test, y_train, y_test = train_test_split(sampled_data, sampled_labels, test_size=0.3, random_state=42)
+
+X_train = pd.DataFrame(X_train)
+X_test = pd.DataFrame(X_test)
+
+sampled_params = {c: list(zip(np.mean(X_train[y_train == c], axis=0), np.std(X_train[y_train == c], axis=0))) for c in classes}
+
+# --- Notre implémentation de Naive Bayes ---
+print("Notre Naive Bayes")
+predicted_nb = predict_bayes_all(X_test, sampled_params)
+print(compute_metrics(y_test, predicted_nb))
+
+# --------------------------------------------
+
+# --- SKLearn Naive Bayes ---
+print("Sklearn Naive Bayes")
+gnb = GaussianNB()
+gnb.fit(X_train, y_train)
+predicted_gnb = gnb.predict(X_test)
+print(compute_metrics(y_test, predicted_gnb))
+
+# ----------------------------
+
+# --- Notre implémentation de Logistic Regression ---
+print("Notre Logistic Regression")
+theta = softmax.train_log_reg_2(X_train, y_train, np.zeros((len(classes), X_train.shape[1] + 1)), 1000, 1e-4)
+predicted_logreg = softmax.predict_log_reg_2(X_test, theta)
+print(compute_metrics(y_test, predicted_logreg))
+
+
+# --------------------------------------------
+
+# --- SKLearn Logistic Regression ---
+print("Sklearn Logistic Regression")
+lr = LogisticRegression()
+lr.fit(X_train, y_train)
+predicted_lr = lr.predict(X_test)
+print(compute_metrics(y_test, predicted_lr))
+
+# --------------------------------------------
+
+# ---------------------------------------------------------------------------
+
+
+
+
+
+
+
+