sklearn封装keras函数:
keras.wrappers.scikit_learnKerasRegressor
keras.wrappers.scikit_learnKerasRegressor函数
思路
RandomizedSearchCV
1.转化为sklearn的model
2.定义参数集合
3.搜索参数
核心
#定义model函数,作为keras.wrappers.scikit_learnKerasRegressor的参数
def build_model(hidden_layers=1,
layer_size=30,
learning_rate=3e-3):
model=keras.models.Sequential()
model.add(keras.layers.Dense(layer_size,activation='relu',input_shape=x_train.shape[1:]))
for i in range(hidden_layers-1):
model.add(keras.layers.Dense(layer_size,activation='relu'))
model.add(keras.layers.Dense(1))
optimizer=keras.optimizers.SGD(learning_rate)
model.compile(loss='mse',optimizer=optimizer)
return model
#keras.wrappers.scikit_learn.KerasRegressor
sklearn_model=keras.wrappers.scikit_learn.KerasRegressor(build_model)
callbacks=[keras.callbacks.EarlyStopping(patience=5,min_delta=1e-2)]
history=sklearn_model.fit(x_train_scaled,y_train,
epochs=100,
validation_data=(x_valid_scaled,y_valid),
callbacks=callbacks)
def plot_learning_curves(history):
pd.DataFrame(history.history).plot(figsize=(8,5))
plt.grid(True)#网格
plt.gca().set_ylim(0,1)#坐标轴范围
plt.show()
plot_learning_curves(history)
#注意!!!
#KerasRegressor' object has no attribute 'evaluate'
'''
test_score=sklearn_model.evaluate(x_test_scaled,y_test)
#打印出每个参数是什么意思
#print(model.metrics_names)
print('{:.2f}'.format(test_score))
自动搜索超参数
from scipy.stats import reciprocal
#f(x)=1/(x*log(b/a)) a<=x<=b
param_distribution={
"hidden_layers":[1,2,3,4],
"layer_size":np.arrange(1,100),
"learning_rate":reciprocal(1e-4,1e-2),
}
from sklearn.model_selection import RandomizedSearchCV
random_search_cv=RandomizedSearchCV(sklearn_model,
param_distribution,
n_iter=10,#从param_distribution里形成多少个参数集合
n_jobs=5#并行处理的个数
)
random_search_cv.fit(x_train_scaled,y_train,epochs=100,
validation_data=(x_valid_scaled,y_valid),
callbacks=callbacks)
#最好的参数
print(random_search_cv.best_params_)
#最好的峰值
print(random_search_cv.best_score_)
#最好的模型
print(random_search_cv.best_estimator_)
model=random_search_cv.best_estimator_.model
model.evaluate(x_test_scaled,y_test)
全部
#RandomizedSearchCV
#1.转化为sklearn的model
#2.定义参数集合
#3.搜索参数
def build_model(hidden_layers=1,
layer_size=30,
learning_rate=3e-3):
model=keras.models.Sequential()
model.add(keras.layers.Dense(layer_size,activation='relu',input_shape=x_train.shape[1:]))
for i in range(hidden_layers-1):
model.add(keras.layers.Dense(layer_size,activation='relu'))
model.add(keras.layers.Dense(1))
optimizer=keras.optimizers.SGD(learning_rate)
model.compile(loss='mse',optimizer=optimizer)
return model
#keras.wrappers.scikit_learn.KerasRegressor
sklearn_model=keras.wrappers.scikit_learn.KerasRegressor(build_model)
callbacks=[keras.callbacks.EarlyStopping(patience=5,min_delta=1e-2)]
history=sklearn_model.fit(x_train_scaled,y_train,
epochs=100,
validation_data=(x_valid_scaled,y_valid),
callbacks=callbacks)
'''
本文介绍如何使用KerasRegressor将Keras模型转换为Scikit-Learn模型,以便利用RandomizedSearchCV进行超参数搜索。通过定义模型构建函数、设置参数分布和执行随机搜索,实现对隐藏层数量、层大小及学习率的有效优化。

被折叠的 条评论
为什么被折叠?



