keras学习笔记（1）

 tf.__version__>>>>>2.2.0

端午节假期在宿舍学keras，实验室跑了一条蛇，结果都没人管，该怎么说呢，环境真的是差到极致。今年的yq，也不能回北京实习，被迫在zh已经呆了一个半月了，吃饭喝水都特别不方便，真的是环境很糟糕，月底学术报告完打算回家，现在也找不到实习，暑假一个人呆在这里，不说每天吃饭点外卖吃到肚子难受，就一个人在这种环境也难受，还是要回家学习。这次回家要安排好计划，每天充实的度过。对于sbrz我好像没那么生气，毕竟我从来没在乎过lj。

import tensorflow as tf
import pandas as pd
print('Tensorflow Version:{}'.format(tf.__version__))
data=pd.read_csv(r'D:DeepLearningTensorflowmy_studyIncome1.csv')
import matplotlib.pyplot as plt
%matplotlib inline
plt.scatter(data.Education,data.Income)
x=data.Education
y=data.Income
model=tf.keras.Sequential()
model.add(tf.keras.layers.Dense(1,input_shape=(1,)))#输出维度是1，输入维度是1,元组的形式,dense层ax+b
model.summary()
model.compile(optimizer='adam',loss='mse')#mse:mean square error
history=model.fit(x,y,epochs=1000)
model.predict(pd.Series([20]))

　　

线性回归

import tensorflow as tf
import pandas as ps
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
data=pd.read_csv(r'D:DeepLearningTensorflowmy_studyAdvertising.csv')
data.head()
plt.scatter(data.TV,data.sales)
plt.scatter(data.radio,data.sales)
plt.scatter(data.newspaper,data.sales)
x=data.iloc[:,1:-1]#数字，取所有行，列第一列到倒数第二列
y=data.iloc[:,-1]
model=tf.keras.Sequential([tf.keras.layers.Dense(10,input_shape=(3,),activation='relu'),tf.keras.layers.Dense(1)])
model.summary()
model.compile(optimizer='adam',loss='mse')
model.fit(x,y,epochs=100)

二分类逻辑回归

import tensorflow as tf
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
data=pd.read_csv(r'D:DeepLearningTensorflowmy_studycredit-a.csv',header=None)#没有表头columns，则header=none
data.head()
data.iloc[:,-1].value_counts()
x=data.iloc[:,:-1]
y=data.iloc[:,-1].replace(-1,0)#把最后一列的-1换成0，这种适合回归模型，-1和1适合支持向量机模型
model=tf.keras.Sequential()
model.add(tf.keras.layers.Dense(4,input_shape=(15,),activation='relu'))
model.add(tf.keras.layers.Dense(4,activation='relu'))
model.add(tf.keras.layers.Dense(1,activation='sigmoid'))
model.summary()
model.compile(optimizer='adam',loss='binary_crossentropy',metrics=['acc'])
history=model.fit(x,y,epochs=100)
history.history.keys()
plt.plot(history.epoch,history.history.get('loss'))
plt.plot(history.epoch,history.history.get('acc'))

多分类逻辑回归

import tensorflow as tf
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
(train_images,train_labels),(test_images,test_labels)=tf.keras.datasets.fashion_mnist.load_data()
train_images.shape,train_labels.shape
plt.imshow(train_images[0])#展示第一幅图
#数据归一化
train_images=train_images/255
test_images=test_images/255
model=tf.keras.Sequential()
model.add(tf.keras.layers.Flatten(input_shape=(28,28)))#变成28*28的一维、
model.add(tf.keras.layers.Dense(128,activation='relu'))
model.add(tf.keras.layers.Dense(10,activation='softmax'))
model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['acc'])
model.fit(train_images,train_labels,epochs=5)
model.evaluate(test_images,test_labels)
#改成独热编码，损失函数要换成categorical_crossentropy
train_label_onehot=tf.keras.utils.to_categorical(train_labels)
test_label_onehot=tf.keras.utils.to_categorical(test_labels)
mdoel=tf.keras.Sequential()
mdoel.add(tf.keras.layers.Flatten(input_shape=(28,28)))
mdoel.add(tf.keras.layers.Dense(128,activation='relu'))
mdoel.add(tf.keras.layers.Dropout(0.5))
model.add(tf.keras.layers.Dense(128,activation='relu'))
model.add(tf.keras.layers.Dropout(0.5))
model.add(tf.keras.layers.Dense(128,activation='relu'))
model.add(tf.keras.layers.Dense(10,activation='softmax'))
model.compile(optimizer=tf.keras.optimizers.Adam(lr=0.01),loss='categorical_crossentropy',metrics=['acc'])
history=model.fit(train_images,train_label_onehot,epochs=5,validation_data=(test_images,test_label_onehot))
history.history.keys()
plt.plot(history.epoch,history.history.get('loss'),label='loss')
plt.plot(history.epoch,history.history.get('val_loss'),label='val_loss')
plt.legend()
plt.plot(history.epoch,history.history.get('acc'),label='acc')
plt.plot(history.epoch,history.history.get('val_acc'),label='val_acc')
plt.legend()
predict=model.predict(test_images)
predict.shape
np.argmax(predict[0])#横向的最大值下标

　函数式API

import tensorflow as tf
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
from tensorflow import keras
(train_images,train_labels),(test_images,test_labels)=tf.keras.datasets.fashion_mnist.load_data()
train_images=train_images/255
test_images=test_images/255
input=keras.Input(shape=(28,28))
x=keras.layers.Flatten()(input)
x=keras.layers.Dense(32,activation='relu')(x)
x=keras.layers.Dropout(0.5)(x)
x=keras.layers.Dense(64,activation='relu')(x)
output=keras.layers.Dense(10,activation='softmax')(x)
model=keras.Model(inputs=input,outputs=output)
model.summary()

　　tf.data模块

import tensorflow as tf
dataset=tf.data.Dataset.from_tensor_slices([1,2,3,4,5,6,7])#数据转换成tensor类型
for ele in dataset:
    print(ele.numpy())#tensor类型转换成numpy类型
for ele in dataset.take(4):#取出前4个
    print(ele)
dataset=tf.data.Dataset.from_tensor_slices([[1,2],[3,4],[5,6]])
for ele in dataset:
    print(ele.numpy())
dataset=tf.data.Dataset.from_tensor_slices({'a':[1,2,3,4],'b':[6,7,8,9],'c':[2,3,4,5]})
for ele in dataset:
    print(ele)
next(iter(dataset.take(1)))#'TakeDataset' object is not an iterator:iter（）可以转换成可迭代对象
#dataset对于数据的变换：
dataset=dataset.shuffle(buffer_size=7)#buffer_size:对多少数据乱序
dataset=dataset.repeat(count=3)#重复shuffle多少次，默认无限,重复3次
data=dataset.batch(3)#每次出来3个数
dataset=tf.data.Dataset.from_tensor_slices([1,2,3,4,5,6,7])
for ele in dataset:
    print(ele.numpy())
dataset=tf.data.Dataset.from_tensor_slices([1,2,3,4,5,6,7])#数据换成tensor类型
#使用函数对数变换
dataset=dataset.map(tf.square)
for ele in dataset:
    print(ele.numpy())
(train_images,train_labels),(test_images,test_labels)=tf.keras.datasets.mnist.load_data()
train_images=train_images/255
test_images=test_images/255
ds_train_img=tf.data.Dataset.from_tensor_slices(train_images)
ds_train_lab=tf.data.Dataset.from_tensor_slices(train_labels)
ds_train=tf.data.Dataset.zip((ds_train_img,ds_train_lab))#zip（）里面是元组类型
ds_train=ds_train.shuffle(10000).repeat().batch(64)#每次
model=tf.keras.Sequential([tf.keras.layers.Flatten(input_shape=(28,28)),tf.keras.layers.Dense(128,activation='relu'),
                          tf.keras.layers.Dense(10,activation='softmax')])
model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['accuracy'])
steps_per_epoch=train_images.shape[0]//64#整除
model.fit(ds_train,epochs=5,steps_per_epoch=steps_per_epoch)
ds_test=tf.data.Dataset.from_tensor_slices((test_images,test_labels))
ds_test=ds_test.batch(64)
validation_steps=10000//64
model.fit(ds_train,epochs=5,steps_per_epoch=steps_per_epoch,validation_data=ds_test,validation_steps=validation_steps)

　　卷积神经网路

import tensorflow as tf
print(tf.__version__)
tf.test.is_gpu_available()#测试gpu是否可用

from tensorflow import keras
import matplotlib.pyplot as plt
%matplotlib inline
import numpy as np
(train_images,train_labels),(test_images,test_labels)=keras.datasets.fashion_mnist.load_data()
train_images.shape
#将一堆图片数据的维度扩张成4维的数据
train_images=np.expand_dims(train_images,-1)

train_images.shape
test_images=np.expand_dims(test_images,-1)
model=tf.keras.Sequential()
model.add(tf.keras.layers.Conv2D(64,(3,3),input_shape=train_images.shape[1:],activation='relu',padding='same'))
model.add(tf.keras.layers.Conv2D(64,(3,3),activation='relu',padding='same'))
model.add(tf.keras.layers.MaxPool2D())#默认2*2
model.add(tf.keras.layers.Dropout(0.5))
model.add(tf.keras.layers.Conv2D(128,(3,3),activation='relu',padding='same'))
model.add(tf.keras.layers.Conv2D(128,(3,3),activation='relu',padding='same'))
model.add(tf.keras.layers.MaxPool2D())#默认2*2
model.add(tf.keras.layers.Dropout(0.5))
model.add(tf.keras.layers.Conv2D(256,(3,3),activation='relu',padding='same'))
model.add(tf.keras.layers.Conv2D(256,(3,3),activation='relu',padding='same'))
model.add(tf.keras.layers.MaxPool2D())#默认2*2
model.add(tf.keras.layers.Dropout(0.5))
model.add(tf.keras.layers.Conv2D(512,(3,3),activation='relu',padding='same'))
model.add(tf.keras.layers.Conv2D(512,(3,3),activation='relu',padding='same'))
model.add(tf.keras.layers.Dropout(0.5))
model.add(tf.keras.layers.GlobalAveragePooling2D())#(None,12,12,64)》》》(None,64),用图片尺寸大小作为池化核大小，即把特征图片划分成1*1大小，现在每张图片提取到64个特征
model.add(tf.keras.layers.Dense(256,activation='softmax'))
model.add(tf.keras.layers.Dense(10,activation='softmax'))
model.summary()
model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['acc'])
history=model.fit(train_images,train_labels,epochs=30,validation_data=(test_images,test_labels))
history.history.keys()
plt.plot(history.epoch,history.history.get('acc'),label='acc')
plt.plot(history.epoch,history.history.get('val_acc'),label='val_acc')
plt.plot(history.epoch,history.history.get('loss'),label='loss')
plt.plot(history.epoch,history.history.get('val_loss'),label='val_loss')