PyTorch Tutorial 08 - Logistic Regres...

教程Python代码如下：

# 1) Design model(input, output size, forward pass)

# 2) Construct loss and optimizer

# 3) Training loop 训练循环

#  - forward pass: compute prediction

#  - backward pass: gradients

#  - update weights

import torch

import torch.nn as nn # 神经网络模块

import numpy as np # 进行数据转换

from sklearn import datasets # 加载二进制分类数据集

from sklearn.preprocessing import StandardScaler # 缩放features

from sklearn.model_selection import train_test_split #将训练集和测试集分离开

"""二进制分类问题，根据输入数据特征，预测乳腺癌"""

# 0) prepare data

bc = datasets.load_breast_cancer() # 乳腺癌数据集

X, y = bc.data, bc.target

n_samples, n_features = X.shape

#print(n_samples, n_features)

"""train_test_split函数用于将矩阵随机划分为训练子集和测试子集，并返回划分好的训练集测试集样本和训练集测试集标签"""

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1234)

# scale 缩放特征

sc = StandardScaler()

X_train = sc.fit_transform(X_train)

X_test = sc.transform(X_test)

X_train = torch.from_numpy(X_train.astype(np.float32))

X_test = torch.from_numpy(X_test.astype(np.float32))

y_train = torch.from_numpy(y_train.astype(np.float32))

y_test = torch.from_numpy(y_test.astype(np.float32))

# 重塑y_train

y_train = y_train.view(y_train.shape[0],1) #现在y_train只有1行

y_test = y_test.view(y_test.shape[0],1)

# 1) model

"""f = wx + b, sigmoid at the end"""

class LogisticRegression(nn.Module):

  def __init__(self,n_input_features):

    super(LogisticRegression, self).__init__()

    self.linear = nn.Linear(n_input_features, 1) #输出仅一个值

  #前向传递，激活函数sigmoid

  def forward(self, x):

    y_predicted = torch.sigmoid(self.linear(x))

    return y_predicted

model = LogisticRegression(n_features)

# 2) loss and optimizer

learning_rate = 0.01

criterion = nn.BCELoss() # 二分类交叉熵损失

optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)

# 3) training loop

num_epoch = 500

for epoch in range(num_epoch):

  # forward pass and loss

  y_predicted = model(X_train)

  loss = criterion(y_predicted,y_train)

  # backward pass

  loss.backward()

  # update weigths

  optimizer.step()

  # zero gradients

  optimizer.zero_grad()

  if(epoch+1) % 10 == 0:

    print(f'epoch: {epoch+1}, loss = {loss.item():.4f}')

# 评估模型

with torch.no_grad():

  y_predicted = model(X_test)

  y_predicted_cls = y_predicted.round()

  acc = y_predicted_cls.eq(y_test).sum() / float(y_test.shape[0])

  print(f'accuracy: {acc:.4f}')