梯度算法求ax+b

目的：

tf.kernel的一般步骤 梯度算法的原理

import tensorflow as tf

print('ts-version:{}'.format(tf.__version__))

ts-version:2.3.0

import pandas as pd

data = pd.read_csv('C:/Users/DengSer/Desktop/brand.csv')

data

	age	maney
0	4	6
1	6	9
2	7	10
3	8	12
4	9	13
5	10	16

import matplotlib.pyplot as plt
%matplotlib inline

plt.scatter(data.age,data.maney)

<matplotlib.collections.PathCollection at 0x1d9f7efe588>

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-fqf8lAPK-1599789210176)(output_6_1.png)]

x=data.age
y=data.maney

model = tf.keras.Sequential()

model.add(tf.keras.layers.Dense(1,input_shape=(1,)))

model.summary() #ax+b

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 1)                 2         
=================================================================
Total params: 2
Trainable params: 2
Non-trainable params: 0
_________________________________________________________________

model.compile(optimizer='adam',#梯度算法模型
             loss='mse'
             )

history= model.fit(x,y,epochs=5000)#训练次数

Epoch 1/5000
1/1 [==============================] - 0s 23ms/step - loss: 422.4163
Epoch 2/5000
1/1 [==============================] - 0s 0s/step - loss: 422.0647
Epoch 3/5000
。
。
。
 Epoch 4996/5000
1/1 [==============================] - 0s 998us/step - loss: 0.8053
Epoch 4997/5000
1/1 [==============================] - 0s 991us/step - loss: 0.8053
Epoch 4998/5000
1/1 [==============================] - 0s 997us/step - loss: 0.8052
Epoch 4999/5000
1/1 [==============================] - 0s 998us/step - loss: 0.8052
Epoch 5000/5000
1/1 [==============================] - 0s 998us/step - loss: 0.8051

model.predict(x)

array([[ 7.111453],
       [ 9.518102],
       [10.721427],
       [11.924751],
       [13.128076],
       [14.3314  ]], dtype=float32)

model.predict(pd.Series([20]))#good good study more more money

array([[26.364645]], dtype=float32)