Learning_rate 不是合法参数

我正在尝试通过实现 GridSearchCV 来测试我的模型。但我似乎无法在 GridSearch 中添加学习率和动量作为参数。每当我尝试通过添加这些代码来执行代码时，我都会收到错误。

这是我创建的模型：

def define_model(optimizers="SGD"):
    model = models.Sequential()
    model.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same', input_shape=(32, 32, 3)))
    model.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))
    model.add(MaxPooling2D((2, 2)))
    model.add(Flatten())
    model.add(Dense(128, activation='relu', kernel_initializer='he_uniform'))
    model.add(Dense(10, activation='softmax'))
    model.compile(loss='binary_crossentropy', optimizer=optimizers, metrics='accuracy')
    return model

我已经实现的 GridSearch：

learn_rate=(0.0001,0.001)
momentum = (0.1, 0.5)
epochs = [5]
batches = [16]

model = KerasClassifier(build_fn=define_model, verbose=2)
param_grid = dict(epochs = epochs, lr = learn_rate, momentum = momentum, batch_size = batches)
grid = GridSearchCV(estimator=model, param_grid= param_grid, n_jobs = 1, cv = 3)

grid_result = grid.fit(trainX, trainY)
print("Best: %f using %s" %(grid_result.best_score_, grid_result.best_params_))

这是我遇到的错误：

~\anaconda3\envs\tf-gpu\lib\site-packages\tensorflow\python\keras\wrappers\scikit_learn.py in check_params(self, params)
    104       else:
    105         if params_name != 'nb_epoch':
--> 106           raise ValueError('{} is not a legal parameter'.format(params_name))
    107 
    108   def get_params(self, **params):  # pylint: disable=unused-argument

ValueError: lr is not a legal parameter

您的参数中没有优化器。因此，您不需要将其作为函数中的参数。相反，您可能需要在函数中提及learning_rate 和momentum 作为参数，并直接在应该的位置添加SGD：

def define_model(lr, momentum):
    model = models.Sequential()
    model.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same', input_shape=(32, 32, 3)))
    model.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))
    model.add(MaxPooling2D((2, 2)))
    model.add(Flatten())
    model.add(Dense(128, activation='relu', kernel_initializer='he_uniform'))
    model.add(Dense(10, activation='softmax'))
    model.compile(loss='binary_crossentropy', optimizer=SGD(lr, momentum), metrics='accuracy')

    return model