【人工智能项目实践】Day1基于逻辑回归的芯片质量检测任务(分类问题) 以及手写数字识别

预备知识

芯片质量检测任务代码详解

核心过程：定、数、模、训、测、上
(问题定义、数据准备、模型构建、模型训练和测试、模型预测、模型部署上线）

1 问题定义

芯片检测数据集（三列：x1,x2,y）是一个分类数据集，适用于逻辑回归模型。
标签y是0或1，代表芯片是否通过检测，x1和x2是芯片的两个评价指标数据。
数据集下载链接：https://download.csdn.net/download/adreammaker/86245857

2 数据准备

将数据进行划分

# 加载深度学习三件套 和 绘图工具
import numpy as np
import pandas as pd
import keras #这里深度学习工具包使用基于tensorflow的keras
from matplotlib import pyplot as plt

data = pd.read_csv('data.csv')
X = data.drop(['y'], axis=1) # 列方向删除
y = data.loc[:, 'y'] # 所有行都要, y列

# 绘图查看图像
#plt.rcParams['font.sans-serif'] = ['Arial Unicode MS'] # 用于mac jupyter
plt.rcParams['font.sans-serif'] = ['simhei'] # 用于linux jupyter
plt.figure(figsize=(5, 5))# 设置画布大小  英寸
passed = plt.scatter(X.loc[:, 'x1'][y==1],X.loc[:, 'x2'][y==1], color='green')
failed = plt.scatter(X.loc[:, 'x1'][y==0],X.loc[:, 'x2'][y==0], color='blue')
plt.legend((passed, failed), ('通过', '不通过'))
plt.show()

可以看出这个数据线性不可分，需要使用激活函数，逻辑回归就是在线性回归模型的基础上加了一个sigmoid激活函数，可以解决非线性分类问题。

# 数据集划分 3:1. 75%:25%
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test  = train_test_split(X, y, random_state=0)

3模型构建

模型准备（小口诀：模型、损失、优化器；）

from keras.models import Sequential
from keras.layers import Dense, Activation # dense就是全链接层
## 网络
mlp = Sequential()
mlp.add(Dense(units = 20, input_dim = 2, activation = 'sigmoid'))
mlp.add(Dense(units=1, activation = 'sigmoid') )
mlp.summary()
## 优化器 损失函数
mlp.compile(optimizer='adam', loss = 'binary_crossentropy')

4模型训练 + 5模型测试

mlp.fit(X_train, y_train, epochs=2000)
pred2 = mlp.predict(X_test)

result = []
for i in range(len(pred2)):
    if pred2[i][0] > 0.5:
        result.append(1)
    else:
        result.append(0)
y_pred = np.array(result)

accuray = sum(y_pred==y_test)/len(y_pred)
accuray

手写数字识别项目

数据可自行下载，也可通过代码自动下载：Downloading data from https://s3.amazonaws.com/img-datasets/mnist.npz

from keras.datasets import mnist
import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
plt.rcParams['font.sans-serif'] = ['Arial Unicode MS']

(X_train, y_train),  (X_test, y_test) = mnist.load_data()
X_train.shape, y_train.shape, X_test.shape, y_test.shape
plt.imshow(X_train[0], cmap='gray')
plt.show()
print(y_train[0])

from keras.models import Sequential
from keras.layers import Dense, Activation
mlp = Sequential()
mlp.add(Dense(units = 392, input_dim = 784, activation = 'relu'))
mlp.add(Dense(units= 392, activation = 'relu') )
mlp.add(Dense(units= 10, activation = 'softmax') )# fc 层
mlp.summary()

mlp.compile(optimizer='adam', loss = 'categorical_crossentropy', metrics=['categorical_accuracy'])
X_train = X_train.reshape(X_train.shape[0], 784)
X_test = X_test.reshape(X_test.shape[0], 784)
# (x - xmin ) / (xmax - xmin ) 把数据 放到一个尺度下 [0, 1]
# 80/100 90/150
X_train = X_train/255
X_test = X_test/255

import tensorflow as tf
y_train = tf.one_hot(indices=y_train, depth=10,on_value=1, off_value=0, axis=1)
y_test = tf.one_hot(indices=y_test, depth=10,on_value=1, off_value=0, axis=1)
mlp.fit(X_train, y_train, epochs=10)

y_pred = mlp.predict(X_test)
y_pred = np.argmax(y_pred, axis=1)
y_pred = tf.convert_to_tensor(y_pred)
mlp.evaluate(X_test, y_test)