3维，梯度下降算法

# author: Roy.G
import numpy as np
import dataset
import plot_utils as pu

def sigmoid(x):
    return 1/(1+np.exp(-x))

xs,ys = dataset.get_beans(100)

w1 = 0.1
w2 = 0.1
b  = 0.1
xs1=xs[:,0]
xs2=xs[:,1]
def forward_propgation (xs1,xs2):
    z=w1*xs1+w2*xs2+b
    a=sigmoid(z)
    return a

for i in range(1000):
        for i in range(100):
            x1 = xs1[i]
            x2 = xs2[i]
            y = ys[i]

            a=forward_propgation(x1,x2)
            e=(y-a)**2

            de_da=-2*(y-a)
            da_dz=a*(1-a)
            dz_dw1=x1
            dz_dw2=x2
            dz_db=1

            de_dw1=de_da*da_dz*dz_dw1
            de_dw2=de_da*da_dz*dz_dw2
            de_db=de_da*da_dz*dz_db

            alpha=0.05
            w1=w1-alpha*de_dw1
            w2 = w2 - alpha * de_dw2
            b = b - alpha * de_db
# pre=forward_propgation(xs1,xs2)
pu.show_scatter_surface(xs,ys,forward_propgation)