线性回归模型
这里注意 tensorflow2.0 默认代的不是完整版 keras
pip install keras -U –pre
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
from tensorflow import keras
print(tf.__version__)
from keras.models import Sequential
from keras.layers import Dense
X=np.linspace(-1,1,200)
np.random.shuffle(X) # 将数据集随机化
Y=0.5*X+2+np.random.normal(0,0.05,(200,))# 假设我们真实模型为：Y=0.5X+2
plt.scatter(X,Y)
plt.show()
X_train,Y_train=X[:160],Y[:160]
X_test,Y_test=X[160:],Y[160:]
model=Sequential()# Keras 有两种类型的模型，序贯模型（Sequential）和函数式模型
# 比较常用的是 Sequential，它是单输入单输出的 model.add(Dense(input_dim=1, units=1))# 通过 add() 方法一层层添加模型
# Dense 是全连接层，第一层需要定义输入，# 第二层无需指定输入，一般第二层把第一层的输出作为输入
定义完模型就需要训练了，不过训练之前我们需要指定一些训练参数
通过 compile() 方法选择损失函数和优化器
这里我们用均方误差作为损失函数，随机梯度下降作为优化方法
model.compile(loss=’mse’,optimizer=’sgd’)
print(“training…”)
for step in range(301):
cost=model.train_on_batch(X_train,Y_train)
print('traincost',cost)
training…
traincost 4.216369
traincost 4.0576506
traincost 3.905112
traincost 3.7585092
traincost 3.6176078
traincost 3.482185
traincost 3.352024
traincost 3.226918
traincost 3.1066682
traincost 2.9910836
traincost 2.8799806
traincost 2.773183
traincost 2.6705215
traincost 2.5718336
traincost 2.4769628
traincost 2.3857594
traincost 2.2980797
traincost 2.2137847
traincost 2.1327415
traincost 2.054823
traincost 1.9799061
traincost 1.9078728
traincost 1.8386104
traincost 1.7720104
traincost 1.7079678
traincost 1.6463829
traincost 1.5871592
traincost 1.5302042
traincost 1.4754289
traincost 1.4227486
traincost 1.3720804
traincost 1.3233463
traincost 1.27647
traincost 1.2313795
traincost 1.1880047
traincost 1.1462784
traincost 1.1061363
traincost 1.0675167
traincost 1.0303605
traincost 0.99460983
traincost 0.9602106
traincost 0.92710984
traincost 0.8952567
traincost 0.86460304
traincost 0.8351019
traincost 0.8067085
traincost 0.77937984
traincost 0.75307435
traincost 0.7277524
traincost 0.70337594
traincost 0.6799079
traincost 0.65731347
traincost 0.63555855
traincost 0.6146107
traincost 0.59443855
traincost 0.5750123
traincost 0.5563029
traincost 0.53828263
traincost 0.520925
traincost 0.5042044
traincost 0.4880963
traincost 0.472577
traincost 0.45762387
traincost 0.44321498
traincost 0.42932963
traincost 0.41594753
traincost 0.40304962
traincost 0.39061695
traincost 0.378632
traincost 0.3670775
traincost 0.35593694
traincost 0.34519458
traincost 0.33483508
traincost 0.3248439
traincost 0.31520706
traincost 0.305911
traincost 0.29694295
traincost 0.2882902
traincost 0.27994087
traincost 0.2718835
traincost 0.26410693
traincost 0.25660056
traincost 0.24935432
traincost 0.2423583
traincost 0.23560306
traincost 0.22907951
traincost 0.22277899
traincost 0.21669313
traincost 0.21081392
traincost 0.20513353
traincost 0.1996446
traincost 0.19434002
traincost 0.18921284
traincost 0.18425646
traincost 0.17946461
traincost 0.17483108
traincost 0.17035016
traincost 0.16601607
traincost 0.16182347
traincost 0.1577671
traincost 0.15384206
traincost 0.15004347
traincost 0.14636669
traincost 0.14280725
traincost 0.1393609
traincost 0.13602349
traincost 0.13279115
traincost 0.12966004
traincost 0.12662646
traincost 0.12368695
traincost 0.12083809
traincost 0.11807661
traincost 0.1153994
traincost 0.11280348
traincost 0.1102859
traincost 0.107843935
traincost 0.105474874
traincost 0.10317614
traincost 0.100945294
traincost 0.098779894
traincost 0.09667765
traincost 0.0946364
traincost 0.09265399
traincost 0.09072839
traincost 0.08885764
traincost 0.08703984
traincost 0.08527314
traincost 0.08355583
traincost 0.08188622
traincost 0.080262646
traincost 0.0786836
traincost 0.077147566
traincost 0.075653054
traincost 0.07419874
traincost 0.07278321
traincost 0.07140521
traincost 0.07006353
traincost 0.068756916
traincost 0.06748423
traincost 0.06624437
traincost 0.065036274
traincost 0.06385889
traincost 0.062711254
traincost 0.06159237
traincost 0.060501337
traincost 0.059437264
traincost 0.05839932
traincost 0.05738666
traincost 0.05639849
traincost 0.055434037
traincost 0.054492556
traincost 0.05357336
traincost 0.052675746
traincost 0.05179907
traincost 0.05094268
traincost 0.050105996
traincost 0.04928838
traincost 0.04848926
traincost 0.047708116
traincost 0.046944384
traincost 0.046197575
traincost 0.045467213
traincost 0.044752765
traincost 0.044053786
traincost 0.04336986
traincost 0.042700525
traincost 0.04204539
traincost 0.041404035
traincost 0.040776096
traincost 0.040161185
traincost 0.03955894
traincost 0.038969003
traincost 0.038391046
traincost 0.037824757
traincost 0.037269793
traincost 0.03672588
traincost 0.036192704
traincost 0.03566999
traincost 0.035157472
traincost 0.034654856
traincost 0.034161918
traincost 0.033678383
traincost 0.033204023
traincost 0.0327386
traincost 0.032281894
traincost 0.03183369
traincost 0.03139376
traincost 0.030961925
traincost 0.030537969
traincost 0.030121703
traincost 0.029712945
traincost 0.029311508
traincost 0.028917229
traincost 0.02852993
traincost 0.028149456
traincost 0.02777564
traincost 0.027408326
traincost 0.027047377
traincost 0.026692647
traincost 0.02634399
traincost 0.026001269
traincost 0.025664348
traincost 0.025333118
traincost 0.025007427
traincost 0.024687175
traincost 0.024372231
traincost 0.024062494
traincost 0.02375785
traincost 0.023458183
traincost 0.023163397
traincost 0.022873381
traincost 0.022588048
traincost 0.022307292
traincost 0.022031024
traincost 0.021759152
traincost 0.021491585
traincost 0.021228248
traincost 0.020969043
traincost 0.020713892
traincost 0.020462718
traincost 0.020215448
traincost 0.019972008
traincost 0.019732315
traincost 0.019496303
traincost 0.019263903
traincost 0.019035045
traincost 0.018809652
traincost 0.018587684
traincost 0.018369064
traincost 0.018153723
traincost 0.017941615
traincost 0.017732676
traincost 0.017526846
traincost 0.017324071
traincost 0.017124299
traincost 0.016927475
traincost 0.016733546
traincost 0.016542466
traincost 0.016354186
traincost 0.016168648
traincost 0.01598582
traincost 0.015805641
traincost 0.015628079
traincost 0.015453073
traincost 0.015280595
traincost 0.0151106
traincost 0.014943043
traincost 0.014777884
traincost 0.014615083
traincost 0.014454606
traincost 0.014296411
traincost 0.014140455
traincost 0.013986712
traincost 0.013835144
traincost 0.013685715
traincost 0.01353839
traincost 0.013393136
traincost 0.013249922
traincost 0.013108711
traincost 0.012969477
traincost 0.01283219
traincost 0.012696815
traincost 0.012563321
traincost 0.012431686
traincost 0.012301879
traincost 0.012173861
traincost 0.012047628
traincost 0.0119231315
traincost 0.011800354
traincost 0.011679266
traincost 0.011559845
traincost 0.011442062
traincost 0.0113259
traincost 0.011211328
traincost 0.011098323
traincost 0.010986872
traincost 0.010876941
traincost 0.0107685095
traincost 0.010661555
traincost 0.010556059
traincost 0.010451991
traincost 0.010349341
traincost 0.0102480855
traincost 0.010148199
traincost 0.010049667
traincost 0.009952474
traincost 0.009856592
traincost 0.009762008
traincost 0.009668704
traincost 0.009576656
traincost 0.009485848
traincost 0.009396264
traincost 0.009307886
traincost 0.009220699
traincost 0.00913468
traincost 0.009049815
traincost 0.008966093
traincost 0.008883496
traincost 0.008802002
traincost 0.0087216
traincost 0.008642277
cost=model.evaluate(X_test,Y_test,batch_size=40)
print(‘test cost’,cost)
W,b = model.layers[0].get_weights()
查看训练出的网络参数
# 由于我们网络只有一层，且每次训练的输入只有一个，输出只有一个
# 因此第一层训练出 Y =WX+ B 这个模型，其中 W,b 为训练出的参数 print(“weights”,W,’n biases’,b)
40/40 [==============================] – 0s 722us/step
test cost 0.007749117910861969
weights [[0.6209358]]
biases [1.9840264]
Y_pre=model.predict(X_test)
plt.scatter(X_test,Y_test)
plt.plot(X_test,Y_pre)
plt.show()