基于numpy用梯度上升法处理逻辑斯蒂回归

import numpy as np 

import matplotlib.pyplot as plt 

w=700

w1=700

n=w1-50

train=np.random.randint(-300,300,(w,4))

train=train.astype(float)

train_lable=np.zeros((w,1))

traint=train.astype(float)

lam=100

for i in range(4):

    train[:,i]=(train[:,i]-train[:,i].mean())/train[:,i].std()

for i in range(w):

    if 1*train[i,0]+2*train[i,1]+3*train[i,2]+4*train[i,3]-1>0:

        train_lable[i]=1

    else:

        train_lable[i]=0

w=np.zeros(4)

b=0

beta=1

for i in range(300):

    sum=np.zeros(5)

    for i1 in range(4):

        for i2 in range(w1-50):

            sum[i1]=sum[i1]+train_lable[i2]*train[i2,i1]-train[i2,i1]*np.exp(np.dot(w,train[i2])+b)/(1+np.exp(np.dot(w,train[i2])+b))-lam/n*w[i1]

    for i2 in range(w1-50):

        sum[4]=sum[4]+train_lable[i2]-np.exp(np.dot(w,train[i2])+b)/(1+np.exp(np.dot(w,train[i2])+b))

    loss=0

    for i2 in range(w1-50):

        loss=loss+train_lable[i2]*(np.dot(w,train[i2])+b)-np.log(1+np.exp(np.dot(w,train[i2])+b))  

    sum=sum/(w1-50)

    loss=loss/(w1-50)-lam/2/n*np.dot(w,w)

    if loss>=-0.9 and beta>=1:

        beta=beta/10

    print(i,beta,loss,sum,w,b)

    for i1 in range(5):

        if i1==4:

            b=beta*sum[4]+b

        else:

            w[i1]=w[i1]+beta*sum[i1]

acc=0

k=w1-50

for i in range(50):

    if(np.dot(w,train[i+k])+b>0):

        if train_lable[i+k]==1:

            acc+=1

    else:

        if train_lable[i+k]==0:

            acc+=1

print(acc)