Keras模型上的递归特征消除(RFE) - TypeError:无法腌制_thread.RLock对象

时间:2018-05-05 00:17:22

标签: python scikit-learn keras feature-selection rfe

我想跟进上一个问题(Recursive Feature Elimination on Keras Models),因为我遇到了障碍。 我目前正在尝试实现以下内容(并非所有代码都在这里以便于阅读):

from keras.wrappers.scikit_learn import KerasClassifier
from sklearn.feature_selection import RFE

k_model  = KerasClassifier(build_fn=model, epochs=epochs, 
batch_size=bs, verbose=0) #model is a standard Keras MLP
selector  = RFE(k_model, step=1)

然而,这似乎是功能性的,在我尝试适合模型的下一行:

selector  = selector.fit(x_train, y_train)

我收到以下错误:

TypeError: can't pickle _thread.RLock objects

任何想法/帮助将不胜感激。 编辑:

# =============================================================================
from keras.models import Sequential
from keras.layers import Dense
from keras.utils import np_utils, plot_model
from sklearn.model_selection import train_test_split
from sklearn.ensemble import AdaBoostClassifier as ABC
from keras.wrappers.scikit_learn import KerasClassifier
from sklearn.feature_selection import RFE
import numpy as np
import time
# =============================================================================
# Functions:
# MLP(x, y)
# =============================================================================
def MLP(x, y):
    s_time = time.clock()
#   ATR, MOM, RSI, OBV
#   keep_features = ['ATR', 'MOM', 'RSI', 'OBV']
#   drop_features = list(set(list(x)).difference(keep_features))
#   x.drop(drop_features, axis=1, inplace=True)    
#    x = x.as_matrix()
#    y = y.as_matrix()
#    num_features = np.array((5,10,15))
    num_classes = 3
    epochs = 1
    bs = 10
    np.random.seed(7)

    x_train,  x_test,  y_train,  y_test  = train_test_split(x, y, test_size=0.33)

#    y_train  = np_utils.to_categorical(y_train, num_classes)
#    y_test   = np_utils.to_categorical(y_test, num_classes)

    # create model
    model  = Sequential()    
    model.add( Dense(50, input_shape=(x_train.shape[1],), activation='tanh'))    
    model.add( Dense(num_classes, activation='softmax'))

    # Compile model
    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
#
# =============================================================================
#     plot_model(model, to_file='../3_Deliverables/Final Paper/data/keras_model.png')
# =============================================================================

    # Port Keras Framework into SK-Learn
    k_model  = KerasClassifier(build_fn=model, epochs=epochs, batch_size=bs, verbose=0)
    temp = k_model
    selector = RFE(temp, step=1)
    out = selector.fit(x_train, y_train)

    # evaluate the model
#    scores = model.evaluate(x_test,  y_test)
#    print("\n%s: %.2f%%" % (indicators.metrics_names[1], scores[1]*100))

    e_time = time.clock()
    print('\n Total Time: ', e_time-s_time)

错误堆栈:

File "C:\Miniconda3\lib\site-packages\spyder\utils\site\sitecustomize.py", 
line 705, in runfile
execfile(filename, namespace)

File "C:\Miniconda3\lib\site-packages\spyder\utils\site\sitecustomize.py", 
line 102, in execfile
exec(compile(f.read(), filename, 'exec'), namespace)

File "C:/Users/Schifman Ben/2_Software/project.py", line 51, in <module>
pred.MLP(x_all, y_all)

File "C:\Users\Schifman Ben\2_Software\prediction.py", line 50, in MLP
out = selector.fit(x_train, y_train)

File "C:\Miniconda3\lib\site-packages\sklearn\feature_selection\rfe.py", 
line 134, in fit
return self._fit(X, y)

File "C:\Miniconda3\lib\site-packages\sklearn\feature_selection\rfe.py", 
line 169, in _fit
estimator = clone(self.estimator)

File "C:\Miniconda3\lib\site-packages\sklearn\base.py", line 62, in clone
new_object_params[name] = clone(param, safe=False)

File "C:\Miniconda3\lib\site-packages\sklearn\base.py", line 53, in clone
return copy.deepcopy(estimator)

File "C:\Miniconda3\lib\copy.py", line 180, in deepcopy
y = _reconstruct(x, memo, *rv)

File "C:\Miniconda3\lib\copy.py", line 280, in _reconstruct
state = deepcopy(state, memo)

File "C:\Miniconda3\lib\copy.py", line 150, in deepcopy
y = copier(x, memo)

File "C:\Miniconda3\lib\copy.py", line 240, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)

File "C:\Miniconda3\lib\copy.py", line 150, in deepcopy
y = copier(x, memo)

File "C:\Miniconda3\lib\copy.py", line 215, in _deepcopy_list
append(deepcopy(a, memo))

File "C:\Miniconda3\lib\copy.py", line 180, in deepcopy
y = _reconstruct(x, memo, *rv)

File "C:\Miniconda3\lib\copy.py", line 280, in _reconstruct
state = deepcopy(state, memo)

File "C:\Miniconda3\lib\copy.py", line 150, in deepcopy
y = copier(x, memo)

File "C:\Miniconda3\lib\copy.py", line 240, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)

File "C:\Miniconda3\lib\copy.py", line 150, in deepcopy
y = copier(x, memo)

File "C:\Miniconda3\lib\copy.py", line 215, in _deepcopy_list
append(deepcopy(a, memo))

File "C:\Miniconda3\lib\copy.py", line 180, in deepcopy
y = _reconstruct(x, memo, *rv)

File "C:\Miniconda3\lib\copy.py", line 280, in _reconstruct
state = deepcopy(state, memo)

File "C:\Miniconda3\lib\copy.py", line 150, in deepcopy
y = copier(x, memo)

File "C:\Miniconda3\lib\copy.py", line 240, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)

File "C:\Miniconda3\lib\copy.py", line 180, in deepcopy
y = _reconstruct(x, memo, *rv)

File "C:\Miniconda3\lib\copy.py", line 280, in _reconstruct
state = deepcopy(state, memo)

File "C:\Miniconda3\lib\copy.py", line 150, in deepcopy
y = copier(x, memo)

File "C:\Miniconda3\lib\copy.py", line 240, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)

File "C:\Miniconda3\lib\copy.py", line 180, in deepcopy
y = _reconstruct(x, memo, *rv)

File "C:\Miniconda3\lib\copy.py", line 280, in _reconstruct
state = deepcopy(state, memo)

File "C:\Miniconda3\lib\copy.py", line 150, in deepcopy
y = copier(x, memo)

File "C:\Miniconda3\lib\copy.py", line 240, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)

File "C:\Miniconda3\lib\copy.py", line 180, in deepcopy
y = _reconstruct(x, memo, *rv)

File "C:\Miniconda3\lib\copy.py", line 280, in _reconstruct
state = deepcopy(state, memo)

File "C:\Miniconda3\lib\copy.py", line 150, in deepcopy
y = copier(x, memo)

File "C:\Miniconda3\lib\copy.py", line 240, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)

File "C:\Miniconda3\lib\copy.py", line 169, in deepcopy
rv = reductor(4)

TypeError: can't pickle _thread.RLock objects

1 个答案:

答案 0 :(得分:0)

KerasClassifier build_fn需要一个指向函数的指针,该函数返回model,而不是模型本身。

所以改变你的代码,如:

def MLP(x, y):

    ...
    ...
    ...
    x_train,  x_test,  y_train,  y_test  = train_test_split(x, y, test_size=0.33)

#    y_train  = np_utils.to_categorical(y_train, num_classes)
#    y_test   = np_utils.to_categorical(y_test, num_classes)

    # This is what you need
    # create model
    def create_model():
        model = Sequential()    
        model.add(Dense(50, input_shape=(x_train.shape[1],), activation='tanh'))    
        model.add(Dense(num_classes, activation='softmax'))

        # Compile model
        model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
        return model

    ...
    ...

    # Port Keras Framework into SK-Learn
    k_model  = KerasClassifier(build_fn=create_model, epochs=epochs, batch_size=bs, verbose=0)
    ...
    ...

但之后你会收到另一个关于feature_importances_的错误,因为RFE与KerasClassifier不兼容。

有关详细信息,请参阅此问题:Feature selection on a keras model