第三个作业

# -*- coding:utf-8 _*-
import random

import numpy as np


def sigmod_Z(X, W_raw):
    # sigmod(Z)
    x_r = np.dot(X, W_raw.T)
    result = 1 / (1 + np.exp(-x_r))
    return result


def deta(sg, X, t):
    # Deta
    dt0 = (sg - t) * sg * (1 - sg)
    dt1 = (sg - t) * sg * (1 - sg) * X[1]
    dt2 = (sg - t) * sg * (1 - sg) * X[2]
    dt = [dt0, dt1, dt2]
    return dt


def y_pre(Xinput, W):
    # 预测值
    sigmod2 = np.dot(Xinput, W.T)
    result = 1 / (1 + np.exp(-sigmod2))
    return result


# ----------------------- 训练 ------------
# 最大损失
Error = 0.001
AF = 0.001

train_data = np.array([[1, 1, 1], [1, 0, 0], [0, 1, 0], [0, 0, 0]])

pre = []
rel = []

j = random.randint(0, 2)
# 输入训练参数
x1 = train_data[j][0]
x2 = train_data[j][1]
t = train_data[j][2]
X = np.array(([1, x1, x2]))
# W初值
w0 = 1
w1 = 1
w2 = 1
W_raw = np.array(([w0, w1, w2]))
print("输入训练的参数X:", X)

dt0 = 0.1
dt1 = 0.1
dt2 = 0.1

while dt0 >= Error or dt1 >= Error or dt2 >= Error:
    w0 = w0 - AF * dt0
    w1 = w1 - AF * dt1
    w2 = w2 - AF * dt2
    W = np.array(([w0, w1, w2]))
    print("----------------------W:", W)
    X = np.array(([1, x1, x2]))
    sg = sigmod_Z(X, W)
    dt = deta(sg, X, t)
    dt0 = dt[0]
    dt1 = dt[1]
    dt2 = dt[2]
    print(dt)

# 最终的参数
W = np.array([w0, w1, w2])
X_input = np.array([0, 1, 0])
y_pre = y_pre(X_input, W)
print("W:", W)
print("X_input", X_input)
print("y_pre:", y_pre)