《PyTorch深度学习实践》Lecture_06 逻辑回归 Logistic Regression

B站刘二大人老师的《PyTorch深度学习实践》Lecture_06 重点回顾+手记+代码复现+报错解决

Lecture_06 逻辑回归 Logistic Regression

一、重点回顾

（一）分类任务的数据集

① The MNIST Dataset：handwritten digits

• Training set: 60,000 examples,
• Test set: 10,000 examples.
• Classes: 10

import torchvision
train_set = torchvision.datasets.MNIST(root='../dataset/mnist', train=True, download=True)
test_set = torchvision.datasets.MNIST(root='../dataset/mnist', train=False, download=True)

② The CIFAR-10 dataset：32×32的彩色小图片

• Training set: 50,000 examples,
• Test set: 10,000 examples.
• Classes: 10（飞机、车、鸟、猫、鹿、狗、青蛙、马、船、卡车）

import torchvision
train_set = torchvision.datasets.CIFAR10(…)
test_set = torchvision.datasets.CIFAR10(…)

（二）Sigmoid functions

S型、饱和函数（导函数的极限为0）

Logistic Function

Logistic函数是最典型的Sigmoid函数

其他的Sigmoid Functions

（三）逻辑回归模型 Logistic Regression Model

import torch.nn.functional as F

class LogisticRegressionModel(torch.nn.Module):
	def __init__(self):
	super(LogisticRegressionModel, self).__init__()
	self.linear = torch.nn.Linear(1, 1)
	
def forward(self, x):
	y_pred = F.sigmoid(self.linear(x))
	return y_pred

criterion = torch.nn.BCELoss()

二、手记

三、代码复现

import torch
import torch.nn.functional as F

# Prepare dataset
x_data = torch.Tensor([[1.0],[2.0],[3.0]])
y_data = torch.Tensor([[0],[0],[1]])

# Design Model
class LogisticRegressionModel(torch.nn.Module):
    def __init__(self):
        super(LogisticRegressionModel,self).__init__()
        self.linear = torch.nn.Linear(1,1)

    def forward(self, x):
        y_pred = F.sigmoid(self.linear(x))
        return  y_pred

model = LogisticRegressionModel()

# Construct Loss & Optimizer
criterion = torch.nn.BCELoss()
optimizer = torch.optim.SGD(model.parameters(),lr=0.01)

# Training Cycle
for epoch in range(1000):
	y_pred = model(x_data)
	loss = criterion(y_pred,y_data)
	print(epoch,loss.item())

	optimizer.zero_grad()
	loss.backward()
	optimizer.step()

# Result of Logistic Regression
import numpy as np
import matplotlib.pyplot as plt

x = np.linspace(0,10,200)
x_t = torch.Tensor(x).view((200,1))
y_t = model(x_t)
y = y_t.data.numpy()
plt.plot(x,y)
plt.plot([0,10],[0.5,0.5],c = 'r')
plt.xlabel('Hours')
plt.ylabel('Probability of Pass')
plt.grid()
plt.show()

注意：torch.sigmoid()、torch.nn.Sigmoid()和torch.nn.functional.sigmoid()三者比较

四、代码报错及解决

OMP: Error #15: Initializing libiomp5md.dll, but found libiomp5md.dll already initialized.

解决：

1. 不完全解决方案

import os
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"

As an unsafe, unsupported, undocumented workaround you can set the environment variable KMP_DUPLICATE_LIB_OK=TRUE to allow the program to continue to execute, but that may cause crashes or silently produce incorrect results.

2. 升级numpy包（完美解决）
   command命令下：pip install --upgrade numpy

参考：
https://blog.csdn.net/qq_37791134/article/details/107912975