机器学习10

单层感知机、多层感知机 推导编码

# import torch
#
# a = torch.zeros(2, 3)
# b = torch.ones(2, 3)
# print(a)
# print(b)
# c = torch.rand(2, 3)
# print(c)
#
# d = torch.where(c > 0.5, a, b)
# # c 中的值大于0.5取a否则取b
# print(d)
# # 梯度是向量组合
# # 梯度方向、大小
# # 找极小值解
# x = torch.linspace(-100, 100, 10)
# print(x)
# print(torch.sigmoid(x))
#
# x = torch.linspace(-1, 1, 10)
# print(x)
# print(torch.tanh(x))
# print(torch.relu(x))

import torch
import torch.nn.functional as F
a = torch.rand(3)
a.requires_grad_()
print(a)
x = torch.randn(1, 10)
print(x)
w = torch.randn(1, 10, requires_grad=True)
# 1行10列二维矩阵
'''单一输出感知机'''
print(w)
print('矩阵')
print(x@w.t())
#  step 1
o = torch.sigmoid(x@w.t())
#  step 2
print(f"-----1x1-----{o}")
print(o.shape)
# 均方损失
loss = F.mse_loss(torch.ones(1, 1), o)
#  step 3
print('----------')
print(loss)
print(torch.ones(1, 1))
print(loss.shape) #loss 标量

loss.backward()
#  step 4
# 向后传播
print(w.grad)
# 求梯度 10个
print()

 

# 多层 感知机
# 多输出
import torch
import torch.nn.functional as F
a = torch.rand(3)
a.requires_grad_()
print(a)
x = torch.randn(1, 10)
print(x)
w = torch.randn(2, 10, requires_grad=True)
# 1行10列二维矩阵
'''单一输出感知机'''
print(w)
print('矩阵')
print(x@w.t())
#  step 1
o = torch.sigmoid(x@w.t())
#  step 2
print(f"-----1x1-----{o}")
print(o.shape)
# 均方损失
loss = F.mse_loss(torch.ones(1, 2), o)
#  step 3
print('----------')
print(loss)
print(torch.ones(1, 1))
print(loss.shape) #loss 标量

loss.backward()
#  step 4
# 向后传播
print(w.grad)
# 求梯度 10个
print()