`import numpy as np
x_train=[[1,2],[1,3],[1,6],[1,9],[1,2],[1,3],[1,6],[1,9]]
y_train=[1,1,0,0,1,1,0,0]
x_train_num=len(x_train)
向量化训练组
x_a=np.array(x_train)
x_v=(x_a.T)
y_v=np.array(y_train)
print(f"x_v={x_v},y_v={y_v}")
学习率和训练次数
stu_rate=0.01
train_item=1000
权重向量化初始化
weight=np.zeros((2,1))
基本偏置，避免input=0,output=0
b=0
激活函数-给网络引入非线性运算
def singmo (z):
a=1/(1+np.exp(-z))
return a
开始训练
for i in range(train_item):
#计算线性输入，向量点积，向量化
Z=np.dot(weight.T,x_v)+b
print(f"Z={Z}")
#反向传播求梯度
#初始wight=0时，A=0.5
A=singmo(Z)
print(f"A={A}")
#计算梯度
#a-y
dz=np.zeros((1,8))
dz=A-y_v
dz=dz.T
print(f"dz={dz}")
#gradient
dw=np.zeros((2,1))
dw=np.array(dw)
dw=(np.dot(x_v,dz))/x_train_num
print(f"gradient={dw}")
db=(np.sum(dz))/x_train_num#更新权重
weight=weight-stu_rate*dw
b=b-stu_rate*db
#检测损失函数
if i % 50 ==0:loss = -np.mean(y_v * np.log(A ) + (1 - y_v) * np.log(1 -A ))print(f"权重向量{weight},损失函数{loss}")
def pre (x,weight,b):
def singmo (z):
a=1/(1+np.exp(-z))
return a
x_test_array = np.array(x).T
Z=np.dot(weight.T,x_test_array)+b
result=singmo(Z)
print(f"Z={result}")
return result
x_test=[[1,5],[1,6]]
y_test=[1,0]
pre_test=pre(x_test,weight,b)
y_test_pre=(pre_test>0.5).astype(int)
print(f"预测结果{y_test_pre}")
`
)
)
