tensorflow2笔记:keras的sklearn包装

预先导入数据

from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler

查看数据样式

housing=fetch_california_housing()
scaler=StandardScaler()
x_data=scaler.fit_transform(housing.data)
x_train_full,x_test,y_train_full,y_test=train_test_split(x_data,housing.target)
x_train,x_valid,y_train,y_valid=train_test_split(x_train_full,y_train_full)
x_train.shape,x_valid.shape
((11610, 8), (3870, 8))
import tensorflow as tf

创建build和compile模型的函数

参数为后面scikit-learn包装类实例化所想要的参数(自己设置),返回编译好的模型

def build_model(n_hidden=2,n_neurons=30,learning_rate=0.1,input_shape=[8]):
    model=tf.keras.models.Sequential()
    model.add(tf.keras.layers.InputLayer(input_shape=input_shape))
    for layer in range(n_hidden):
        model.add(tf.keras.layers.Dense(n_neurons,activation='relu',kernel_initializer='he_normal'))
    model.add(tf.keras.layers.Dense(1))
    optimizer=tf.keras.optimizers.SGD(learning_rate=learning_rate,momentum=0.9,nesterov=True)
    model.compile(loss='mse',optimizer=optimizer)
    return model

scikit-learn包装

#scikit-learn包装
keras_reg=tf.keras.wrappers.scikit_learn.KerasRegressor(build_model)

模型训练

注意fit方法会把参数传递到内部的model,就像model.fit()
EarlyStopping:提前停止,防止过拟合,patience为损失或精度停止轮次阈值,restore_best_weights设置停止后返回最好的权重

#提前停止,防止过拟合
earlystop=tf.keras.callbacks.EarlyStopping(patience=10,restore_best_weights=True)
keras_reg.fit(x_train,y_train,epochs=100,validation_data=(x_valid,y_valid),callbacks=[earlystop])
keras_reg.score(x_test,y_test)
Epoch 1/100
363/363 [==============================] - 3s 4ms/step - loss: 1.7924 - val_loss: 1.3685
Epoch 2/100
363/363 [==============================] - 1s 4ms/step - loss: 1.3301 - val_loss: 1.3647
Epoch 3/100
363/363 [==============================] - 1s 4ms/step - loss: 1.3344 - val_loss: 1.3750
Epoch 4/100
363/363 [==============================] - 1s 4ms/step - loss: 1.3357 - val_loss: 1.3937
Epoch 5/100
363/363 [==============================] - 1s 4ms/step - loss: 1.3349 - val_loss: 1.3666
Epoch 6/100
363/363 [==============================] - 3s 7ms/step - loss: 1.3334 - val_loss: 1.4313
Epoch 7/100
363/363 [==============================] - 3s 9ms/step - loss: 1.3374 - val_loss: 1.3662
Epoch 8/100
363/363 [==============================] - 3s 9ms/step - loss: 1.3325 - val_loss: 1.3770
Epoch 9/100
363/363 [==============================] - 5s 14ms/step - loss: 1.3330 - val_loss: 1.4188
Epoch 10/100
363/363 [==============================] - 3s 9ms/step - loss: 1.3321 - val_loss: 1.3701
Epoch 11/100
363/363 [==============================] - 3s 9ms/step - loss: 1.3315 - val_loss: 1.3959
Epoch 12/100
363/363 [==============================] - 5s 14ms/step - loss: 1.3353 - val_loss: 1.3920
162/162 [==============================] - 0s 2ms/step - loss: 1.3511





-1.3510503768920898

用GridSearchCV或RandomizedSearchCV搜索参数空间

注意:建议使用RandomizedSearchCV

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

起名大废废

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付:¥1
获取中
扫码支付

您的余额不足,请更换扫码支付或充值

打赏作者

实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值