Tensorflow2 keras 分类模型

from pickletools import optimize
from pyexpat import model
from re import X
from tkinter import Y
import matplotlib as mpl
import matplotlib.pyplot as plt

from sklearn.preprocessing import StandardScaler
import numpy as np
import sklearn
import pandas as pd
import os
import sys
import time
from sklearn import metrics
import tensorflow as tf

from tensorflow import keras

#数据集
fashion_mnist = keras.datasets.fashion_mnist
#训练集和测试集
(x_train_all,y_train_all),(x_test,y_test) = fashion_mnist.load_data()

#验证集和训练集
x_valid,x_train = x_train_all[:5000],x_train_all[5000:]
y_valid,y_train = y_train_all[:5000],y_train_all[5000:]

# 训练集归一化
# x = （x - u）/ std ：x - 均值 / 方差 
scaler = StandardScaler()

x_train_scaled = scaler.fit_transform(
    x_train.astype(np.float32).reshape(-1,1)).reshape(-1,28,28)

x_valid_scaled = scaler.transform(x_valid.astype(np.float32).reshape(-1,1)).reshape(-1,28,28)
x_test_scaled = scaler.transform(x_test.astype(np.float32).reshape(-1,1)).reshape(-1,28,28)


def show_single_image(img_arr):
    plt.imshow(img_arr,cmap="binary")
    plt.show()

def show_imgs(n_rows,n_cols,x_data,y_data,class_names):
    assert len(x_data) == len(y_data)
    assert n_rows * n_cols < len(x_data)
    #指定图像宽高 英尺单位
    plt.figure(figsize=(n_cols * 1.4,n_rows * 1.6))
    for row in range(n_rows):
        for col in range(n_cols):        
            index = n_cols * row + col
            #创建单个子图
            plt.subplot(n_rows,n_cols,index + 1)
            plt.imshow(x_data[index],cmap="binary",interpolation='nearest')
            plt.axis('off')
            plt.title(class_names[y_data[index]])
    plt.show()

class_names = ['T-shirt','Trouser','Pullover','Dress','Coat','Sandal','Shirt','Sneaker','Bag','Ankle boot']



#show_imgs(3,5,x_train,y_train,class_names)


#添加模型 sequential线性堆叠模型
model = keras.models.Sequential()
#将28*28的矩阵展平为一维向量
model.add(keras.layers.Flatten(input_shape=[28,28]))
#Dense：每一层的输入来自前面所有层的输出->解决梯度消失的问题
#梯度消失和梯度爆炸：计算深度增加导致求导数据持续过低（0-0.25）或过高（1）
model.add(keras.layers.Dense(300,activation="relu"))
#此100单元与300单元做全联接
#relu：y = max(0,x) 大于0返回x
#softmax：将向量变成概率分布 x = [x1,x2,x3]
# y = [e^x1/sum, e^x2/sum,e^x3/sum] sum = e^x1/sum+e^x2/sum+e^x3/sum

model.add(keras.layers.Dense(100,activation="relu"))
model.add(keras.layers.Dense(10,activation="softmax"))

# sparse_categorical_crossentropy: y是一个数值需要将 y->one_hot->[] 转化为向量,如果是向量需要用categorical_crossentropy
# optimize 模型调整方法
# metrics
# optimizer="adam" sgd ->梯度优化算法 
model.compile(loss="sparse_categorical_crossentropy",optimizer="adam", metrics = ["accuracy"])

#模型架构显示
#架构参数：
#1层 [None,784] [样本数*784]
#2层 第一层转化为 [None,300] ：[none,784] * w + b -> [none,300] w.shape[784,300], b=[300]
model.summary()

#结果验证 
history = model.fit(x_train_scaled,y_train,epochs=10,validation_data=(x_valid_scaled,y_valid))
# history.history

#结果估值
print(model.evaluate(x_test_scaled,y_test))

#结果可视化
def plot_learning_curves(history):
    pd.DataFrame(history.history).plot(figsize=(8,5))
    plt.grid(True)
    plt.gca().set_ylim(0,1)
    plt.show()
plot_learning_curves(history)