人工智能概论第三次作业

import torch
import torch.nn as nn
import torch.optim as optim
import numpy as np
import matplotlib.pyplot as plt
% matplotlib
inline

# 生成数据
num_samples = 300
num_classes = 3
feature_dim = 2

np.random.seed(1)
data = np.zeros((num_samples, feature_dim))
labels = np.zeros(num_samples, dtype=int)

for i in range(num_samples):
    if i % 3 == 0:
        data[i] = np.random.randn(feature_dim) * 0.1 + np.array([0, -0.5])
        labels[i] = 0
    elif i % 3 == 1:
        data[i] = np.random.randn(feature_dim) * 0.1 + np.array([-0.5, 0.5])
        labels[i] = 1
    else:
        data[i] = np.random.randn(feature_dim) * 0.1 + np.array([0.5, 0.5])
        labels[i] = 2

# 数据可视化
plt.scatter(data[:, 0], data[:, 1], c=labels, cmap='rainbow')
plt.show()

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.fc1 = nn.Linear(2, 10)
        self.fc2 = nn.Linear(10, 3)

    def forward(self, x):
        x = self.fc1(x)
        x = nn.functional.relu(x)
        x = self.fc2(x)
        return x

net = Net()

criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.1)

num_epochs = 50

for epoch in range(num_epochs):
    running_loss = 0.0

    for i in range(num_samples):
        inputs = torch.tensor(data[i], dtype=torch.float32).unsqueeze(0)
        labels_ = torch.tensor(labels[i], dtype=torch.long).unsqueeze(0)

        optimizer.zero_grad()

        outputs = net(inputs)
        loss = criterion(outputs, labels_)
        loss.backward()
        optimizer.step()

        running_loss += loss.item()

    print('Epoch %3d | Loss: %.4f' % (epoch + 1, running_loss / num_samples))

print('Finished Training')

# 计算模型预测结果
predicts = np.zeros(num_samples, dtype=int)

for i in range(num_samples):
    inputs = torch.tensor(data[i], dtype=torch.float32).unsqueeze(0)
    outputs = net(inputs)
    predicts[i] = outputs.argmax().item()

# 绘制分类结果
plt.scatter(data[:, 0], data[:, 1], c=predicts, cmap='rainbow')
plt.show()

#生成的分类结果如下图所示：

#![result](https: // img - blog.csdnimg.cn / 2021040515030366.
#png)