PyTorch学习（13）——自编码（AutoEncoder）

本篇博客主要介绍PyTorch中的自编码（AutoEncoder），并使用自编码来实现非监督学习。

示例代码：

import torch
import torch.nn as nn
import torch.utils.data as Data
import torchvision
from torch.autograd import Variable
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
from matplotlib import cm
import numpy as np

# 超参数
EPOCH = 10
BATCH_SIZE = 64
LR = 0.005
DOWNLOAD_MNIST = False
N_TEST_IMG = 5

# 下载MNIST数据
train_data = torchvision.datasets.MNIST(
    root='./mnist/',
    train=True,
    transform=torchvision.transforms.ToTensor(),
    download=DOWNLOAD_MNIST,
)

# 输出一个样本
# print(train_data.train_data.size())
# print(train_data.train_labels.size())
# plt.imshow(train_data.train_data[2].numpy(), cmap='gray')
# plt.title('%i' % train_data.train_labels[2])
# plt.show()

# Dataloader
train_loader = Data.DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True)


class AutoEncoder(nn.Module):
    def __init__(self):
        super(AutoEncoder, self).__init__()
        self.encoder = nn.Sequential(
            nn.Linear(28 * 28, 128),
            nn.Tanh(),
            nn.Linear(128, 64),
            nn.Tanh(),
            nn.Linear(64, 12),
            nn.Tanh(),
            nn.Linear(12, 3),
        )

        self.decoder = nn.Sequential(
            nn.Linear(3, 12),
            nn.Tanh(),
            nn.Linear(12, 64),
            nn.Tanh(),
            nn.Linear(64, 128),
            nn.Tanh(),
            nn.Linear(128, 28 * 28),
            nn.Sigmoid(),
        )

    def forward(self, x):
        encoded = self.encoder(x)
        decoded = self.decoder(encoded)
        return encoded, decoded

autoencoder = AutoEncoder()
optimizer = torch.optim.Adam(autoencoder.parameters(), lr=LR)
loss_func = nn.MSELoss()

# initialize figure
f, a = plt.subplots(2, N_TEST_IMG, figsize=(5, 2))
plt.ion()   # continuously plot

# original data (first row) for viewing
view_data = train_data.train_data[:N_TEST_IMG].view(-1, 28*28).type(torch.FloatTensor)/255.
for i in range(N_TEST_IMG):
    a[0][i].imshow(np.reshape(Variable(view_data).data.numpy()[i], (28, 28)), cmap='gray'); a[0][i].set_xticks(()); a[0][i].set_yticks(())


for epoch in range(EPOCH):
    for step, (x, y) in enumerate(train_loader):
        b_x = Variable(x.view(-1, 28 * 28))
        b_y = Variable(x.view(-1, 28 * 28))
        b_label = Variable(y)

        encoded, decoded = autoencoder(b_x)

        loss = loss_func(decoded, b_y)
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        if step % 100 == 0:
            print('Epoch: ', epoch, '| train loss: %.4f' % loss.data.numpy())

            # plotting decoded image (second row)
            _, decoded_data = autoencoder(Variable(view_data))
            for i in range(N_TEST_IMG):
                a[1][i].clear()
                a[1][i].imshow(np.reshape(decoded_data.data.numpy()[i], (28, 28)), cmap='gray')
                a[1][i].set_xticks(());
                a[1][i].set_yticks(())
            plt.draw();
            plt.pause(0.05)

plt.ioff()
plt.show()

# visualize in 3D plot
view_data = train_data.train_data[:200].view(-1, 28 * 28).type(torch.FloatTensor) / 255.
encoded_data, _ = autoencoder(Variable(view_data))
fig = plt.figure(2);
ax = Axes3D(fig)
X, Y, Z = encoded_data.data[:, 0].numpy(), encoded_data.data[:, 1].numpy(), encoded_data.data[:, 2].numpy()

values = train_data.train_labels[:200].numpy()
for x, y, z, s in zip(X, Y, Z, values):
    c = cm.rainbow(int(255 * s / 9));
    ax.text(x, y, z, s, backgroundcolor=c)
ax.set_xlim(X.min(), X.max());
ax.set_ylim(Y.min(), Y.max());
ax.set_zlim(Z.min(), Z.max())
plt.show()

数据示例：

运行结果：

Epoch:  0 | train loss: 0.2318
Epoch:  0 | train loss: 0.0704
Epoch:  0 | train loss: 0.0680
Epoch:  0 | train loss: 0.0617
Epoch:  0 | train loss: 0.0595
Epoch:  0 | train loss: 0.0512
Epoch:  0 | train loss: 0.0528
Epoch:  0 | train loss: 0.0486
Epoch:  0 | train loss: 0.0498
Epoch:  0 | train loss: 0.0445
Epoch:  1 | train loss: 0.0468
Epoch:  1 | train loss: 0.0459
Epoch:  1 | train loss: 0.0428
Epoch:  1 | train loss: 0.0447
Epoch:  1 | train loss: 0.0455
Epoch:  1 | train loss: 0.0448
Epoch:  1 | train loss: 0.0422
Epoch:  1 | train loss: 0.0489
Epoch:  1 | train loss: 0.0426
Epoch:  1 | train loss: 0.0417
Epoch:  2 | train loss: 0.0470
Epoch:  2 | train loss: 0.0413
Epoch:  2 | train loss: 0.0398
Epoch:  2 | train loss: 0.0419
Epoch:  2 | train loss: 0.0424
Epoch:  2 | train loss: 0.0418
Epoch:  2 | train loss: 0.0426
Epoch:  2 | train loss: 0.0398
Epoch:  2 | train loss: 0.0401
Epoch:  2 | train loss: 0.0401
Epoch:  3 | train loss: 0.0420
Epoch:  3 | train loss: 0.0444
Epoch:  3 | train loss: 0.0396
Epoch:  3 | train loss: 0.0447
Epoch:  3 | train loss: 0.0367
Epoch:  3 | train loss: 0.0384
Epoch:  3 | train loss: 0.0446
Epoch:  3 | train loss: 0.0435
Epoch:  3 | train loss: 0.0434
Epoch:  3 | train loss: 0.0406
Epoch:  4 | train loss: 0.0379
Epoch:  4 | train loss: 0.0382
Epoch:  4 | train loss: 0.0403
Epoch:  4 | train loss: 0.0351
Epoch:  4 | train loss: 0.0377
Epoch:  4 | train loss: 0.0367
Epoch:  4 | train loss: 0.0370
Epoch:  4 | train loss: 0.0397
Epoch:  4 | train loss: 0.0376
Epoch:  4 | train loss: 0.0353
Epoch:  5 | train loss: 0.0402
Epoch:  5 | train loss: 0.0368
Epoch:  5 | train loss: 0.0382
Epoch:  5 | train loss: 0.0395
Epoch:  5 | train loss: 0.0396
Epoch:  5 | train loss: 0.0414
Epoch:  5 | train loss: 0.0373
Epoch:  5 | train loss: 0.0388
Epoch:  5 | train loss: 0.0363
Epoch:  5 | train loss: 0.0382
Epoch:  6 | train loss: 0.0366
Epoch:  6 | train loss: 0.0357
Epoch:  6 | train loss: 0.0360
Epoch:  6 | train loss: 0.0397
Epoch:  6 | train loss: 0.0376
Epoch:  6 | train loss: 0.0364
Epoch:  6 | train loss: 0.0370
Epoch:  6 | train loss: 0.0383
Epoch:  6 | train loss: 0.0360
Epoch:  6 | train loss: 0.0334
Epoch:  7 | train loss: 0.0369
Epoch:  7 | train loss: 0.0324
Epoch:  7 | train loss: 0.0372
Epoch:  7 | train loss: 0.0373
Epoch:  7 | train loss: 0.0360
Epoch:  7 | train loss: 0.0347
Epoch:  7 | train loss: 0.0352
Epoch:  7 | train loss: 0.0322
Epoch:  7 | train loss: 0.0346
Epoch:  7 | train loss: 0.0359
Epoch:  8 | train loss: 0.0354
Epoch:  8 | train loss: 0.0353
Epoch:  8 | train loss: 0.0345
Epoch:  8 | train loss: 0.0303
Epoch:  8 | train loss: 0.0367
Epoch:  8 | train loss: 0.0356
Epoch:  8 | train loss: 0.0364
Epoch:  8 | train loss: 0.0373
Epoch:  8 | train loss: 0.0375
Epoch:  8 | train loss: 0.0357
Epoch:  9 | train loss: 0.0314
Epoch:  9 | train loss: 0.0355
Epoch:  9 | train loss: 0.0375
Epoch:  9 | train loss: 0.0367
Epoch:  9 | train loss: 0.0364
Epoch:  9 | train loss: 0.0347
Epoch:  9 | train loss: 0.0348
Epoch:  9 | train loss: 0.0333
Epoch:  9 | train loss: 0.0395
Epoch:  9 | train loss: 0.0393