机器学习-识别手写数字0-9

import os
os.environ['TF_CPP_MIN_LOG_LEVEL']='2' # to hidden the messages from tensorflow
from tensorflow import keras
from  tensorflow.keras import layers
import numpy as np
import matplotlib.pyplot as plt
# from tensorflow.keras.datasets import mnist
# mnist is the handwriting number dataset 0-9
import sys

def load_mnist(path):
    file=np.load(path)
    x_train,y_train=file['x_train'],file['y_train']
    x_test,y_test=file['x_test'],file['y_test']
    file.close()
    return (x_train, y_train), (x_test, y_test)


def check(N,pages):
    idx=0
    for page in range(pages):
        for i in range(N):
            for j in range(N):
                num = i*N+j
                plt.subplot(N,N,num+1)
                plt.imshow(x_train[num+idx],cmap=plt.get_cmap('gray'))
        idx+=N*N
        plt.show()
        # show the plot


path="C:/Users/77007/Desktop/python/pythonProject1/mnist.npz"
(x_train,y_train),(x_test, y_test)=load_mnist(path)
# print(x_train.shape) # 60000 张图片 pix 28*28
# print(y_train.shape) # 60000 个结果对应0-9其中一个
x_train=x_train.reshape(-1,784).astype("float32")/255.0
x_test=x_test.reshape(-1,784).astype("float32")/255.0
# print(x_train.shape)
# print(x_test.shape)

# Sequential API (convenient, not flexible)
model=keras.Sequential(
    [
        keras.Input(shape=(28*28)), # for print the model
        layers.Dense(512,activation='relu'),
        layers.Dense(256,activation='relu'),
        layers.Dense(10),
    ]
)
# another definition method
''' 
model=keras.Sequential()
model.add(keras.Input(shape=784))
model.add(layers.Dense(512,activation='relu'))
model.add(layers.Dense(256,activation='relu'))
model.add(layers.Dense(10))
'''
# print(model.summary())
# sys.exit()

# Functional API (more flexible)
inputs=keras.Input(shape=(784))
x=layers.Dense(512,activation='relu',name='first_layer')(inputs)
x=layers.Dense(256,activation='relu',name='second_layer')(x)
outputs=layers.Dense(10,activation='softmax')(x)
model=keras.Model(inputs=inputs,outputs=outputs) # change the model (OR you can comment this line using the model upward)
# print(model.summary())
# for printing summary of our model, we can add name='xxx' feature of each layers

model.compile(
    loss=keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    # Sequential API using True, Functional API using False
    optimizer=keras.optimizers.Adam(lr=0.001),
    # learing rate=0.001
    metrics=["accuracy"],
)

model.fit(x_train,y_train,batch_size=32,epochs=5,verbose=2)
model.evaluate(x_test,y_test,batch_size=32,verbose=2)

Train on 60000 samples
Epoch 1/5
60000/60000 - 6s - loss: 0.1865 - accuracy: 0.9425
Epoch 2/5
60000/60000 - 5s - loss: 0.0800 - accuracy: 0.9749
Epoch 3/5
60000/60000 - 5s - loss: 0.0541 - accuracy: 0.9826
Epoch 4/5
60000/60000 - 5s - loss: 0.0395 - accuracy: 0.9872
Epoch 5/5
60000/60000 - 5s - loss: 0.0341 - accuracy: 0.9890
10000/1 - 0s - loss: 0.0392 - accuracy: 0.9792