域名续费后网站打不开,无锡专业做网站的公司哪家好,wordpress 首行缩进,怎么免费注册自己的网站#x1f368; 本文为#x1f517;365天深度学习训练营 中的学习记录博客 #x1f366; 参考文章#xff1a;365天深度学习训练营-第G7周#xff1a;Semi-Supervised GAN 理论与实战#xff08;训练营内部成员可读#xff09; #x1f356; 原作者#xff1a;K同学啊|接… 本文为365天深度学习训练营 中的学习记录博客 参考文章365天深度学习训练营-第G7周Semi-Supervised GAN 理论与实战训练营内部成员可读 原作者K同学啊|接辅导、项目定制 运行环境 电脑系统Windows 10 语言环境python 3.10 编译器Pycharm 2022.1.1深度学习环境Pytorch 目录
一、理论知识讲解
二、代码实现
1、配置代码 2、初始化权重
3、定义算法模型
4、配置模型 5、训练模型 一、理论知识讲解
该算法将产生式对抗网络(GAN) 拓展到半监督学习通过强制判别器D来输出类别标签。我们 在一个数据集上训练一个生成器G以及一个判别器D输入是N类当中的一个。在训练的时候判别器D被用于预测输入是属于N1类中的哪一个,这个N1是对应了生成器G的输出这里的判别器D同时也充当起了分类器C的效果。这种方法可以用于训练效果更好的判别器D并且可以比普通的GAN产性更加高质量的样本。Semi-Supervised GAN有如下优点: 1作者对GANs做了一个新的扩展允许它同时学习一个生成模型和一个分类器。我们把这个 扩展叫做半监督GAN或SGAN 2论文实验结果表明SGAN在有限数据集比没有生成部分的基准分类器提升了分类性能。 3论文实验结果表明SGAN可以显著地提升生成样本的质量并降低生成器的训练时间。
二、代码实现
1、配置代码
import argparse
import os
import numpy as np
import mathimport torchvision.transforms as transforms
from torchvision.utils import save_imagefrom torch.utils.data import DataLoader
from torchvision import datasets
from torch.autograd import Variableimport torch.nn as nn
import torch.nn.functional as F
import torchos.makedirs(images, exist_okTrue)parser argparse.ArgumentParser()
parser.add_argument(--n_epochs, typeint, default2, helpnumber of epochs of training)
parser.add_argument(--batch_size, typeint, default64, helpsize of the batches)
parser.add_argument(--lr, typefloat, default0.0002, helpadam: learning rate)
parser.add_argument(--b1, typefloat, default0.5, helpadam: decay of first order momentum of gradient)
parser.add_argument(--b2, typefloat, default0.999, helpadam: decay of first order momentum of gradient)
parser.add_argument(--n_cpu, typeint, default2, helpnumber of cpu threads to use during batch generation)
parser.add_argument(--latent_dim, typeint, default100, helpdimensionality of the latent space)
parser.add_argument(--num_classes, typeint, default10, helpnumber of classes for dataset)
parser.add_argument(--img_size, typeint, default32, helpsize of each image dimension)
parser.add_argument(--channels, typeint, default1, helpnumber of image channels)
parser.add_argument(--sample_interval, typeint, default400, helpinterval between image sampling)
opt parser.parse_args(args[])
print(opt)cuda True if torch.cuda.is_available() else False Namespace(n_epochs2, batch_size64, lr0.0002, b10.5, b20.999, n_cpu2, latent_dim100, num_classes10, img_size32, channels1, sample_interval400) 2、初始化权重
def weights_init_normal(m):classname m.__class__.__name__if classname.find(Conv) ! -1:torch.nn.init.normal_(m.weight.data, 0.0, 0.02)elif classname.find(BatchNorm) ! -1:torch.nn.init.normal_(m.weight.data, 1.0, 0.02)torch.nn.init.constant_(m.bias.data, 0.0)
3、定义算法模型
class Generator(nn.Module):def __init__(self):super(Generator, self).__init__()self.label_emb nn.Embedding(opt.num_classes, opt.latent_dim)self.init_size opt.img_size // 4 # Initial size before upsamplingself.l1 nn.Sequential(nn.Linear(opt.latent_dim, 128 * self.init_size ** 2))self.conv_blocks nn.Sequential(nn.BatchNorm2d(128),nn.Upsample(scale_factor2),nn.Conv2d(128, 128, 3, stride1, padding1),nn.BatchNorm2d(128, 0.8),nn.LeakyReLU(0.2, inplaceTrue),nn.Upsample(scale_factor2),nn.Conv2d(128, 64, 3, stride1, padding1),nn.BatchNorm2d(64, 0.8),nn.LeakyReLU(0.2, inplaceTrue),nn.Conv2d(64, opt.channels, 3, stride1, padding1),nn.Tanh(),)def forward(self, noise):out self.l1(noise)out out.view(out.shape[0], 128, self.init_size, self.init_size)img self.conv_blocks(out)return imgclass Discriminator(nn.Module):def __init__(self):super(Discriminator, self).__init__()def discriminator_block(in_filters, out_filters, bnTrue):Returns layers of each discriminator blockblock [nn.Conv2d(in_filters, out_filters, 3, 2, 1), nn.LeakyReLU(0.2, inplaceTrue), nn.Dropout2d(0.25)]if bn:block.append(nn.BatchNorm2d(out_filters, 0.8))return blockself.conv_blocks nn.Sequential(*discriminator_block(opt.channels, 16, bnFalse),*discriminator_block(16, 32),*discriminator_block(32, 64),*discriminator_block(64, 128),)# The height and width of downsampled imageds_size opt.img_size // 2 ** 4# Output layersself.adv_layer nn.Sequential(nn.Linear(128 * ds_size ** 2, 1), nn.Sigmoid())self.aux_layer nn.Sequential(nn.Linear(128 * ds_size ** 2, opt.num_classes 1), nn.Softmax())def forward(self, img):out self.conv_blocks(img)out out.view(out.shape[0], -1)validity self.adv_layer(out)label self.aux_layer(out)return validity, label
4、配置模型
# Loss functions
adversarial_loss torch.nn.BCELoss()
auxiliary_loss torch.nn.CrossEntropyLoss()# Initialize generator and discriminator
generator Generator()
discriminator Discriminator()if cuda:generator.cuda()discriminator.cuda()adversarial_loss.cuda()auxiliary_loss.cuda()# Initialize weights
generator.apply(weights_init_normal)
discriminator.apply(weights_init_normal)# Configure data loader
os.makedirs(../../data/mnist, exist_okTrue)
dataloader torch.utils.data.DataLoader(datasets.MNIST(../../data/mnist,trainTrue,downloadTrue,transformtransforms.Compose([transforms.Resize(opt.img_size), transforms.ToTensor(), transforms.Normalize([0.5], [0.5])]),),batch_sizeopt.batch_size,shuffleTrue,
)# Optimizers
optimizer_G torch.optim.Adam(generator.parameters(), lropt.lr, betas(opt.b1, opt.b2))
optimizer_D torch.optim.Adam(discriminator.parameters(), lropt.lr, betas(opt.b1, opt.b2))FloatTensor torch.cuda.FloatTensor if cuda else torch.FloatTensor
LongTensor torch.cuda.LongTensor if cuda else torch.LongTensor Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ../../data/mnist\MNIST\raw\train-images-idx3-ubyte.gzExtracting ../../data/mnist\MNIST\raw\train-images-idx3-ubyte.gz to ../../data/mnist\MNIST\rawDownloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to ../../data/mnist\MNIST\raw\train-labels-idx1-ubyte.gzExtracting ../../data/mnist\MNIST\raw\train-labels-idx1-ubyte.gz to ../../data/mnist\MNIST\rawDownloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to ../../data/mnist\MNIST\raw\t10k-images-idx3-ubyte.gzExtracting ../../data/mnist\MNIST\raw\t10k-images-idx3-ubyte.gz to ../../data/mnist\MNIST\rawDownloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to ../../data/mnist\MNIST\raw\t10k-labels-idx1-ubyte.gzExtracting ../../data/mnist\MNIST\raw\t10k-labels-idx1-ubyte.gz to ../../data/mnist\MNIST\raw 5、训练模型
# ----------
# Training
# ----------for epoch in range(opt.n_epochs):for i, (imgs, labels) in enumerate(dataloader):batch_size imgs.shape[0]# Adversarial ground truthsvalid Variable(FloatTensor(batch_size, 1).fill_(1.0), requires_gradFalse)fake Variable(FloatTensor(batch_size, 1).fill_(0.0), requires_gradFalse)fake_aux_gt Variable(LongTensor(batch_size).fill_(opt.num_classes), requires_gradFalse)# Configure inputreal_imgs Variable(imgs.type(FloatTensor))labels Variable(labels.type(LongTensor))# -----------------# Train Generator# -----------------optimizer_G.zero_grad()# Sample noise and labels as generator inputz Variable(FloatTensor(np.random.normal(0, 1, (batch_size, opt.latent_dim))))# Generate a batch of imagesgen_imgs generator(z)# Loss measures generators ability to fool the discriminatorvalidity, _ discriminator(gen_imgs)g_loss adversarial_loss(validity, valid)g_loss.backward()optimizer_G.step()# ---------------------# Train Discriminator# ---------------------optimizer_D.zero_grad()# Loss for real imagesreal_pred, real_aux discriminator(real_imgs)d_real_loss (adversarial_loss(real_pred, valid) auxiliary_loss(real_aux, labels)) / 2# Loss for fake imagesfake_pred, fake_aux discriminator(gen_imgs.detach())d_fake_loss (adversarial_loss(fake_pred, fake) auxiliary_loss(fake_aux, fake_aux_gt)) / 2# Total discriminator lossd_loss (d_real_loss d_fake_loss) / 2# Calculate discriminator accuracypred np.concatenate([real_aux.data.cpu().numpy(), fake_aux.data.cpu().numpy()], axis0)gt np.concatenate([labels.data.cpu().numpy(), fake_aux_gt.data.cpu().numpy()], axis0)d_acc np.mean(np.argmax(pred, axis1) gt)d_loss.backward()optimizer_D.step()batches_done epoch * len(dataloader) iif batches_done % opt.sample_interval 0:save_image(gen_imgs.data[:25], images/%d.png % batches_done, nrow5, normalizeTrue)print([Epoch %d/%d] [Batch %d/%d] [D loss: %f, acc: %d%%] [G loss: %f]% (epoch, opt.n_epochs, i, len(dataloader), d_loss.item(), 100 * d_acc, g_loss.item()))[Epoch 0/2] [Batch 937/938] [D loss: 1.358861, acc: 50%] [G loss: 0.671799]
[Epoch 1/2] [Batch 937/938] [D loss: 1.343094, acc: 50%] [G loss: 0.681119]
