02_函数生成权重矩阵

最新推荐文章于 2024-07-12 16:16:27 发布

zhm678

最新推荐文章于 2024-07-12 16:16:27 发布

阅读量38

点赞数

分类专栏： python手搓神经网络文章标签： python

本文链接：https://blog.csdn.net/zhm678/article/details/134077339

版权

python手搓神经网络专栏收录该内容

5 篇文章 0 订阅

订阅专栏

import numpy as np
a1 = -0.9
a2 = 0.5
a3 = 0.7
input1 = np.array([a1, a2, a3])
input1 = np.array([[a1, a2, a3],
                   [-a1, -a2, -a3]])

w11 = 0.8
w21 = -0.4
w31 = 0

w12 = 0.5
w22 = -0.2
w32 = 0

weights = np.array([[w11, w12],
                  [w21, w22],
                  [w31, w32]])

# weights = np.array([w1,
#                     w2,
#                     w3])

b1 = np.array([-100, 100])

# sum1 = a1*w1 + a2*w2 + a3*w3 + b1
# sum1 = np.dot(input1, weights) + b1
# 激活函数
def activation_relu(input):
    return np.maximum(0, input)

# 权重生成函数
def create_weights(n_inputs, n_neurons):
    return np.random.randn(n_inputs, n_neurons)
# 偏置生成函数
def create_bias(n_neurons):
    return np.random.randn(n_neurons)

weights = create_weights(3, 2)
bias = create_bias(2)
print(bias)
sum0 = np.dot(input1, weights)
sum1 = np.dot(input1, weights) + bias

sum2 = activation_relu(sum1)
print(sum0)
print(sum1) 
print(sum2)