def fun(x, y):
# 隐藏层
if Wx1 and Wx2:
b1 = -s1 * Wx1
b2 = -s2 * Wx2
if Wy1 and Wy2:
b1 = -s1 * Wy1
b2 = -s2 * Wy2
if Wx3 and Wx3:
b3 = -s3 * Wx3
b4 = -s4 * Wx4
if Wy4 and Wy4:
b3 = -s3 * Wy3
b4 = -s4 * Wy4
a1 = network2.sigmoid(x * Wx1 + y * Wy1 + b1)
a2 = network2.sigmoid(x * Wx2 + y * Wy2 + b2)
a3 = network2.sigmoid(x * Wx3 + y * Wy3 + b3)
a4 = network2.sigmoid(x * Wx4 + y * Wy4 + b4)
# 输出层
w1 = w3 = h
w2 = w4 = -w1
return network2.sigmoid(w1 * a1 + w2 * a2 + w3 * a3 + w4 * a4 + b)
def n_last_layer_out_put(n, fun):
"""画出最后一层的输出，fun是r-sigmod(f(x))，最终输出将是f(x)
"""
y = None
x = None
for i in np.linspace(0, 1, n, endpoint=False):
s1 = i
s2 = i + 1.0 / n
h = fun(i) # 区间高度
x, y1 = pair_hidden_neurons_out_put(s1, s2, h)
if y == None:
y = y1
else:
y += y1
y = network2.sigmoid(y)
plot_figure.plot_base(y_coordinate=[y],
x_coordinate=[x],
title='n = %d' % (n),
x_lable='X',
y_lable='Sigmod (z)',
x_limit=[min(x) - 0.2,
max(x) + 0.2],
y_limit=[min(y) - 0.2,
max(y) + 0.2])
def fun(x, y):
# 隐藏层
if Wx1 and Wx2:
b1 = -s1 * Wx1
b2 = -s2 * Wx2
if Wy1 and Wy2:
b1 = -s1 * Wy1
b2 = -s2 * Wy2
a1 = network2.sigmoid(x * Wx1 + y * Wy1 + b1)
a2 = network2.sigmoid(x * Wx2 + y * Wy2 + b2)
# 输出层
w1 = h
w2 = -w1
b = 0
return w1 * a1 + w2 * a2 + b
def fun(x, y):
b = -s * w1
return network2.sigmoid(x * w1 + y * w2 + b)
def getOutputOfHiddenUnits(data, net):
output = []
for x,y in data:
output.append(network2.sigmoid(np.dot(net.weights[0], x) + net.biases[0]))
return output
def display_weight(matrix, i):
display(network2.sigmoid(matrix[i]) )
def get_sigmod_coordinate(w, b):
x = np.linspace(0, 1, 10000, endpoint=False) # 输入从0-1，一共10000个点
z = w * x + b
y = network2.sigmoid(z)
return x, y