随机搜索获取参数未实现

Posted 2023-03-12

【中文标题】随机搜索获取参数未实现

【英文标题】：Randomized Search Get param not implemented

【发布时间】：2019-06-19 08:19:19

【问题描述】：

我正在一些图像上训练我的 cnn 模型，并希望添加随机搜索以优化超参数，但我在使用超参数的随机搜索时遇到了麻烦。我正在分享我的模型和一些代码以及我遇到的错误。

我尝试了 sklearn 文档示例和其他文章，但没有从那里获得任何帮助。还尝试了网格搜索，但遇到了同样的问题

def Model(Classes):
    activation='relu' # or linear
    dropout_rate=0.0 # or 0.2
    init_mode='uniform'
    weight_constraint=0 # or  4
    optimizer='adam' # or SGD
    lr = 0.01
    momemntum=0

    #model
    model=Sequential()
    model.add(Conv2D(20,(2,2),padding="same"))
    model.add(Activation(activation))
    model.add(Conv2D(20,(2,2),padding="same"))
    model.add(Activation(activation))
    model.add(MaxPooling2D(pool_size=(2,2),strides=(2,2)))



    model.add(Conv2D(80,(5,5),padding="same"))
    model.add(Activation(activation))
    model.add(Conv2D(80,(5,5),padding="same"))
    model.add(Activation(activation))
    model.add(MaxPooling2D(pool_size=(2,2),strides=(2,2)))



    model.add(Conv2D(100,(5,5),padding="same"))
    model.add(Activation(activation))
    model.add(Conv2D(100,(5,5),padding="same"))
    model.add(Activation(activation))
    model.add(MaxPooling2D(pool_size=(5,5),strides=(2,2)))
    model.add(Dropout(dropout_rate))


    model.add(Flatten())

    model.add(Dense(units=64,activation=activation, 
    kernel_regularizer=regularizers.l2(0.001),
                activity_regularizer=regularizers.l1(0.001)))

    model.add(Dense(units=128,activation=activation, 
    kernel_regularizer=regularizers.l2(0.001),
                activity_regularizer=regularizers.l1(0.01)))

    model.add(Dense(units=Classes,activation=activation))

    model.compile(loss=losses.binary_crossentropy,
                 optimizer=optimizer,
                 metrics=['accuracy'])
    return model




model=Model(2)

activation =  ['relu', 'tanh', 'sigmoid', 'hard_sigmoid', 'linear'] 
optimizer=['adam','adagrad']
epochs = [1, 10] # add 50, 100, 150 etc
batch_size = [1000, 5000] # add 5, 10, 20, 40, 60, 80, 100 etc
param_grid = dict(epochs=epochs, batch_size=batch_size)
#param_grid=dict(epochs=epochs, activation=activation , 
optimizer=optimizer)

random_search = RandomizedSearchCV(model, param_distributions=param_grid,
                               n_iter=10, cv=5,scoring='accuracy')


random_search.fit(X_train, y_train)

report(random_search.cv_results_)

错误：

---------------------------------------------------------------------------  

TypeError                                 Traceback (most recent call last)

      <ipython-input-96-b29b0e381ee4> in <module>()
      12 
      13   ---> 14 random_search.fit(X_train, y_train)
      15 
      16 report(random_search.cv_results_)

      /opt/conda/lib/python3.6/site- 
      packages/sklearn/model_selection/_search.py in fit(self, X, y, 
      groups, **fit_params)
      675         n_splits = cv.get_n_splits(X, y, groups)
      676 
      -> 677         base_estimator = clone(self.estimator)
      678 
      679         parallel = Parallel(n_jobs=self.n_jobs, 
      verbose=self.verbose,

      /opt/conda/lib/python3.6/site-packages/sklearn/base.py in 
      clone(estimator, safe)
      56                             "it does not seem to be a scikit-learn 
      estimator "
      57                             "as it does not implement a 
      'get_params' methods."
      ---> 58                             % (repr(estimator), 
      type(estimator)))
      59     klass = estimator.__class__
      60     new_object_params = estimator.get_params(deep=False)

      TypeError: Cannot clone object '<keras.engine.sequential.Sequential 
      object at 0x7fc268d8abe0>' (type <class 
      'keras.engine.sequential.Sequential'>): it does not seem to be a 
       scikit-learn estimator as it does not implement a 'get_params' 
      methods.

请帮我解决这个问题

【参考方案1】：

你需要像下面这样包装你的模型。

from keras.wrappers.scikit_learn import KerasClassifier
k_model = KerasClassifier(model, verbose=0)

random_search = RandomizedSearchCV(k_model, param_distributions=param_grid,
                                   n_iter=10, cv=5,scoring='accuracy')

你可能想看看this reproducible example。