【PyTorch学习笔记】7.自编码器

淡淡的烟草味﹌ 2022-09-05 13:56 315阅读 0赞

### 文章目录 ###

*  47.Auto-Encoder介绍
 *  48.Auto-Encoder变种
 *  49.Variational Auto-Encoder引入
 *  50.变分自编码器VAE
 *  51.实战
 *   *  51.1AE
     *  51.2VAE

根据龙良曲Pytorch学习视频整理，视频链接：  
[【计算机-AI】PyTorch学这个就够了！][-AI_PyTorch]  
[(好课推荐)深度学习与PyTorch入门实战——主讲人龙良曲][PyTorch]

# 47.Auto-Encoder介绍 #

*  Supervised Learning
 *  Unsupervised Learning
 *  Reinforcement Learning

Why need Unsupervised Learning?

*  Dimension reduction
 *  Preprocessing
 *  Visualization [https://projector.tensorflow.org/][https_projector.tensorflow.org]
 *  Taking advantages of unsupervised data
 *  Compression, denoising, super-resolution…

How to Train?

*  Loss function for binary inputs  
     l ( f ( x ) ) = − ∑ k ( x k l o g ( x ^ k ) + ( 1 − x k ) l o g ( 1 − x ^ k ) ) l(f(x))=-\\sum\_k(x\_klog(\\hat\{x\}\_k)+(1-x\_k)log(1-\\hat\{x\}\_k)) l(f(x))=−∑k(xklog(x^k)\+(1−xk)log(1−x^k))  
    Cross-entropy error function (reconstruction loss)  f ( x ) ≡ x ^ f(x)\\equiv \\hat\{x\} f(x)≡x^
 *  Loss function for real-valued inputs  
     l ( f ( x ) ) = 1 2 ∑ k ( x ^ k − x k ) 2 l(f(x))=\\frac\{1\}\{2\}\\sum\_k(\\hat\{x\}\_k-x\_k)^2 l(f(x))=21∑k(x^k−xk)2  
    sum of squared differences (reconstruction loss)  
    we use a linear activation function at the output

自编码器是一种无监督的数据维度压缩和数据特征表达方法。自编码器是神经网络的一种，由编码器和解码器组成，经过训练后能尝试将输入复制到输出。

PCA V.S. Auto-Encoders

*  PCA, which finds the directions of maximal variance in high-dimensional data, select only those axes that have the largest variance.
 *  The linearity of PCA, however, places significant limitations on the kinds of feature dimensions that can be extracted.

# 48.Auto-Encoder变种 #

Denosing AutoEncoders  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70]  
Dropout AutoEncoders  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 1]  
Adversarial AutoEncoders

*  Distribution of hidden code
 *  Give more details after GAN  
    ![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 2]

# 49.Variational Auto-Encoder引入 #

Another Approach:  q ( z ) → p ( z ) q(z)\\rightarrow p(z) q(z)→p(z)

*  Explicitly enforce  
     l i ( θ , ϕ ) = − E z ∼ q θ ( z ∣ x i ) \[ l o g   p ϕ ( x i ∣ z ) \] + K L ( q θ ( z ∣ x i ) ∣ ∣ p ( z ) ) l\_i(\\theta, \\phi)=-E\_\{z\\sim q\_\{\\theta\}(z|x\_i)\}\[log\\space p\_\{\\phi\}(x\_i|z)\]+KL(q\_\{\\theta\}(z|x\_i)||p(z)) li(θ,ϕ)=−Ez∼qθ(z∣xi)\[log pϕ(xi∣z)\]\+KL(qθ(z∣xi)∣∣p(z))

How to compute KL between  q ( z ) q(z) q(z) and  p ( z ) p(z) p(z)  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 3]  
Sample() is not differentiable  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 4]  
Reparameterization trick  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 5]  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 6]

# 50.变分自编码器VAE #

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 7]

# 51.实战 #

**main.py**

import torch
    from torch.utils.data import DataLoader
    from torchvision import transforms, datasets
    from Pytorch21_7_29.autoencoder.ae import AE
    from Pytorch21_7_29.autoencoder.vae import VAE
    from torch import nn, optim
    import visdom
    
    def main():
        mnist_train = datasets.MNIST('mnist', True, transform=transforms.Compose([
            transforms.ToTensor()
        ]), download=True)
        mnist_train = DataLoader(mnist_train, batch_size=32, shuffle=True)
    
        mnist_test = datasets.MNIST('mnist', True, transform=transforms.Compose([
            transforms.ToTensor()
        ]), download=True)
        mnist_test = DataLoader(mnist_test,batch_size=32,shuffle=True)
    
        x, _ = iter(mnist_train).next()
        print('x:', x.shape)    # x: torch.Size([32, 1, 28, 28])
    
        device = torch.device('cuda')
        # model = AE().to(device)
        model = VAE().to(device)
        criteon = nn.MSELoss()
        optimizer = optim.Adam(model.parameters(), lr=1e-3)
        print(model)
    
        viz = visdom.Visdom()
        for epoch in range(1000):
            for batchidx, (x, _) in enumerate(mnist_train):
                # [b, 1, 28, 28]
                x = x.to(device)
    
                # x_hat = model(x)
                x_hat, kld = model(x)
                loss = criteon(x_hat, x)
    
                if kld is not None:
                    # elbo = -loss - 1.0 * kld
                    # loss = -elbo
                    loss = loss + kld
    
                # backprop
                optimizer.zero_grad()
                loss.backward()
                optimizer.step()
    
            # print(epoch, 'loss', loss.item())
            print(epoch, 'loss', loss.item(), 'kld:', kld.item())
    
            x, _ = iter(mnist_test).next()
            x = x.to(device)
            with torch.no_grad():
                # x_hat = model(x)
                x_hat, _ = model(x)
            viz.images(x, nrow=8, win='x', opts=dict(title='x'))
            viz.images(x_hat, nrow=8, win='x_hat', opts=dict(title='x_hat'))
    
    if __name__ == '__main__':
        main()

## 51.1AE ##

**ae.py**

from torch import nn
    
    class AE(nn.Module):
        def __init__(self):
            super(AE, self).__init__()
    
            # [b, 784] => [b, 20]
            self.encoder = nn.Sequential(
                nn.Linear(784, 256),
                nn.ReLU(),
                nn.Linear(256, 64),
                nn.ReLU(),
                nn.Linear(64, 20),
                nn.ReLU(),
            )
    
            # [b, 20] => [b, 784]
            self.decoder = nn.Sequential(
                nn.Linear(20, 64),
                nn.ReLU(),
                nn.Linear(64, 256),
                nn.ReLU(),
                nn.Linear(256, 784),
                nn.Sigmoid()
            )
    
        def forward(self, x):
            """ :param x: [b, 1, 28, 28] :return: """
            batchsz = x.size(0)
            # flatten
            x = x.view(batchsz, 784)
            # encoder
            x = self.encoder(x)
            # decoder
            x = self.decoder(x)
            # reshape
            x = x.view(batchsz, 1, 28, 28)
            return x

![在这里插入图片描述][f6770cdc555645f48d7c81c8133a7c44.png]

## 51.2VAE ##

**vae.py**

import torch
    from torch import nn
    
    class VAE(nn.Module):
        def __init__(self):
            super(VAE, self).__init__()
    
            # [b, 784] => [b, 20]
            # u: [b, 10] sigma: [b, 10]
            self.encoder = nn.Sequential(
                nn.Linear(784, 256),
                nn.ReLU(),
                nn.Linear(256, 64),
                nn.ReLU(),
                nn.Linear(64, 20),
                nn.ReLU(),
            )
    
            # [b, 10] => [b, 784]
            self.decoder = nn.Sequential(
                nn.Linear(10, 64),
                nn.ReLU(),
                nn.Linear(64, 256),
                nn.ReLU(),
                nn.Linear(256, 784),
                nn.Sigmoid()
            )
    
        def forward(self, x):
            """ :param x: [b, 1, 28, 28] :return: """
            batchsz = x.size(0)
            # flatten
            x = x.view(batchsz, 784)
            # encoder
            # [b, 20], including mean and sigma
            h_ = self.encoder(x)
            # [b, 20] => [b, 10] + [b, 10]
            mu, sigma = h_.chunk(2, dim=1)
            # reparameterization trick, epision~N(0, 1)
            h = mu + sigma * torch.randn_like(sigma)
    
            # decoder
            x_hat = self.decoder(h)
            # reshape
            x_hat = x_hat.view(batchsz, 1, 28, 28)
    
            kld = 0.5 * torch.sum(
                torch.pow(mu, 2) +
                torch.pow(sigma, 2) +
                torch.log(1e-8 + torch.pow(sigma, 2)) - 1
            ) / (batchsz * 28 * 28)
    
            return x_hat, kld

![在这里插入图片描述][913cdfbe7b2b4f2f9bcce8436f7bd954.png]

[-AI_PyTorch]: https://www.bilibili.com/video/BV1Rv411y7oE?p=1
[PyTorch]: https://www.bilibili.com/video/BV1mh41167e3?p=1
[https_projector.tensorflow.org]: https://projector.tensorflow.org/
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70]: /images/20220829/46b161eae90a47bb9d7c1a5499a3d7ff.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 1]: /images/20220829/a308352b100f46a98ca5a1dcb5caffe7.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 2]: /images/20220829/b38121799120472c81c9ba54fb5c912c.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 3]: /images/20220829/d1d2d20e3ebd4fd5a74f09d6aca4575c.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 4]: /images/20220829/b3d3f399f6de4189980a163ab7568c26.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 5]: /images/20220829/69a9c1c525384043976f137ca0224a0f.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 6]: /images/20220829/91efb05217144c70a0605118a00f26f4.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2tlaXZlbl8_size_16_color_FFFFFF_t_70 7]: /images/20220829/f7dbb6cd8ef2406ca5c4754a26e964f3.png
[f6770cdc555645f48d7c81c8133a7c44.png]: /images/20220829/22022a28c39542b2b3ad2e5317715983.png
[913cdfbe7b2b4f2f9bcce8436f7bd954.png]: /images/20220829/325c16d9a2504910818c54e35f7055a9.png