def RandomForest(X, y, epsilon, num_learners, depth, min_sample):
dt = DecisionTree(epsilon, min_sample, depth)
result = Ensemble(dt, num_learners, 0.8)
result.fit(X, y)
return result
print('- Dataset: %d samples' % total_size)
print('- Training set: %d samples' % train_size)
print('- Test set: %d samples' % test_size)
np.random.seed(54)
rp = np.random.permutation(total_size)
idx_train = rp[:train_size]
idx_test = rp[train_size:]
# Datos a usar para entrenamiento y verificacion
X_train, y_train = X[idx_train], y[idx_train]
X_test, y_test = X[idx_test], y[idx_test]
# =============================================================================
# Entrenamos el modelo
# =============================================================================
# Con una GPU GTX1060 este procedimiento toma ~1s por época
Test.fit(X_train,
y_train, (X_test, y_test),
bootstrap_percent=.9,
batch_size=128,
epochs=15)
# =============================================================================
# Cargar el modelo colectivo con componentes entrenadas en clasificadores/
# =============================================================================
Test = Ensemble.load_model('clasificadores')
