Ошибка при запуске байесовского метода поиска лучшего гиперпараметра ⇐ Python

[Anonymous]

Я пытаюсь применить глубокое обучение для создания регрессии (6 независимых переменных и 1 зависимая переменная)
Я пытаюсь использовать байесовскую оптимизацию, чтобы найти лучший параметр для регрессии ИНС (искусственная нейронная сеть), но получаю ошибка.

Код: Выделить всё

# Step 1: Import necessary libraries for Bayesian Optimization
import numpy as np
import pandas as pd
from keras.models import Sequential
from keras.layers import Dense, Dropout, Input
from keras.optimizers import Adam, SGD
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from scikeras.wrappers import KerasRegressor
from sklearn.metrics import mean_absolute_error
from keras.callbacks import EarlyStopping
import warnings
from skopt import BayesSearchCV
from skopt.space import Real, Categorical, Integer

# Suppress warnings for cleaner output
warnings.filterwarnings('ignore')

# Step 2: Load the dataset
df = pd.read_excel('...')

# Define features (X) and target (y)
X = df.drop(columns=['Y'])  # Features
y = df['Y']                 # Target variable

# Step 3: Split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Step 4: Standardize the features and target variable
scaler_X = StandardScaler()
scaler_y = StandardScaler()

# Scale the features
X_train_scaled = scaler_X.fit_transform(X_train)
X_test_scaled = scaler_X.transform(X_test)

# Scale the target variable
y_train_scaled = scaler_y.fit_transform(y_train.values.reshape(-1, 1)).flatten()
y_test_scaled = scaler_y.transform(y_test.values.reshape(-1, 1)).flatten()

# Step 5: Define the model-building function
def make_regression_ann(Optimizer_trial='adam', hidden_layers=1, dropout_rate=0.2,
weight_init='he_normal', activation_function='relu',
learning_rate=0.001, momentum=0.9):
# Choose optimizer based on parameters
if Optimizer_trial == 'adam':
optimizer = Adam(learning_rate=learning_rate)
elif Optimizer_trial == 'sgd':
optimizer = SGD(learning_rate=learning_rate, momentum=momentum)
else:
raise ValueError(f"Unknown optimizer: {Optimizer_trial}")

model = Sequential()

# Input layer + first hidden layer
model.add(Input(shape=(X_train_scaled.shape[1],)))
model.add(Dense(units=5, kernel_initializer=weight_init, activation=activation_function))
model.add(Dropout(rate=dropout_rate))  # Apply dropout

# Additional hidden layers
for _ in range(hidden_layers - 1):
model.add(Dense(units=5, kernel_initializer=weight_init, activation=activation_function))
model.add(Dropout(rate=dropout_rate))  # Apply dropout to each hidden layer

# Output layer
model.add(Dense(1, kernel_initializer='normal'))  # Single output variable

# Compile the model
model.compile(loss='mean_absolute_error', optimizer=optimizer)
return model

# Step 6: Define hyperparameter space for Bayesian Optimization
bayes_search_space = {
'batch_size': Integer(10, 50),  # Integer range for batch size
'epochs': Integer(10, 50),     # Integer range for epochs
'model__Optimizer_trial': Categorical(['adam', 'sgd']),
'model__hidden_layers': Integer(1, 3),  # Number of hidden layers
'model__dropout_rate': Real(0.1, 0.5),  # Dropout rate as a continuous variable
'model__weight_init': Categorical(['he_normal', 'glorot_uniform']),
'model__activation_function': Categorical(['relu', 'tanh']),
'model__learning_rate': Real(1e-4, 1e-2, prior='log-uniform'),  # Learning rate in log scale
'model__momentum': Real(0.5, 0.9)  # Momentum for SGD
}

# Step 7: Wrap the model for Bayesian Optimization
RegModel_Bayes = KerasRegressor(model=make_regression_ann, verbose=0)

# Step 8: Set up Bayesian Optimization
# Adding Early Stopping as a callback
early_stopping_bayes = EarlyStopping(monitor='loss', patience=3, verbose=1)

bayes_search = BayesSearchCV(
estimator=RegModel_Bayes,
search_spaces=bayes_search_space,
n_iter=2,  # Number of iterations for Bayesian Optimization
scoring='neg_mean_absolute_error',
cv=5,  # Reduce CV folds to speed up the process
n_jobs=-1,  # Use all available CPU cores
verbose=0,
random_state=42
)

# Step 9:  Fit the Bayesian Optimization model
try:
bayes_search.fit(X_train_scaled, y_train_scaled, callback=[early_stopping_bayes])
except Exception as e:
print("Bayesian Optimization failed due to memory or worker issues. Consider reducing the parameter space or using fewer iterations.")
print("Error:", e)

# Step 10: Print the best parameters
if bayes_search.best_params_:
print("Best Parameters (Bayesian Optimization):", bayes_search.best_params_)

# Step 11: Evaluate the best model
def evaluate_model(grid_search, X_test_scaled, y_test, scaler_y):
best_model = grid_search.best_estimator_  # Extract the best Keras model
predictions_scaled = best_model.predict(X_test_scaled)  # Make predictions on scaled test data
predictions = scaler_y.inverse_transform(predictions_scaled.reshape(-1, 1))  # Inverse transform predictions
y_test_original = scaler_y.inverse_transform(y_test.reshape(-1, 1))  # Inverse transform y_test
mae = mean_absolute_error(y_test_original, predictions)  # Calculate MAE
print("Mean Absolute Error (MAE) on Test Data:", mae)
return mae

# Step 12: Evaluate the best model
if bayes_search.best_params_:
evaluate_model(bayes_search, X_test_scaled, y_test_scaled, scaler_y)

Но я получаю эту ошибку, но не знаю, как ее исправить. Я попробовалchatgpt, но не могу

Код: Выделить всё

Bayesian Optimization failed due to memory or worker issues. Consider reducing the parameter space or using fewer iterations.
Error: callback should be either a callable or a list of callables.
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[3], line 111
108     print("Error:", e)
110 # Step 10: Print the best parameters
--> 111 if bayes_search.best_params_:
112     print("Best Parameters (Bayesian Optimization):", bayes_search.best_params_)
114 # Step 11: Evaluate the best model

AttributeError: 'BayesSearchCV' object has no attribute 'best_params_'

Как это решить?

Подробнее здесь: https://stackoverflow.com/questions/792 ... rparameter