1 神经网络核心组件

1. 层：神经网络的基本结构，将输入张量转换为输出张量

2. 模型：层构成的网络

3. 损失函数：参数学习的目标函数，通过最小化损失函数来学习各种参数

4. 优化器：如何是损失函数最小

2 神经网络实例

如果初学者，建议直接看3，避免运行结果有误。

2.1 背景说明

2.2 准备数据

##（1）导入必要的模块
import numpy as np
import torch
# 导入内置的 mnist数据
from torchvision.datasets import mnist
# 导入预处理模块
import torchvision.transforms as transforms
from torch.utils.data import DataLoader
# 导入nn及优化器
import torch.nn.functional as F
import torch.optim as optim
from torch import nn
## (2) 定义一些超参数
train_batch_size = 64
test_batch_size = 128
learning_rate = 0.01
num_epoches = 20
lr = 0.01
momentum = 0.5
## （3） 下载数据并对数据进行预处理
# 定义预处理函数，这些预处理依次放在Compose函数中
transform = transforms.Compose([transforms.ToTensor(),transforms.Normalize([0.5],[0.5])])
# 下载数据，并对数据进行预处理
train_dataset = mnist.MNIST('./data', train=True, transform=transform, download=True)
test_dataset = mnist.MNIST('./data', train=False, transform=transform)
# dataloader是一个可迭代的对象，可以使用迭代器一样使用
train_loader = DataLoader(train_dataset, batch_size=train_batch_size, shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=test_batch_size, shuffle=False)

       

2.3 可视化数据源

import matplotlib.pyplot as plt
%matplotlib inline
examples = enumerate(test_loader)
batch_idx, (example_data, example_targets) = next(examples)
fig = plt.figure()
for i in range(6):
  plt.subplot(2,3,i+1)
  plt.tight_layout()
  plt.imshow(example_data[i][0], cmap='gray', interpolation='none')
  plt.title("Ground Truth: {}".format(example_targets[i]))
  plt.xticks([])
  plt.yticks([])

     

2.4 构建模型

## （1）构建网络
class Net(nn.Module):
    """
    使用sequential构建网络，Sequential（）函数功能是将网络的层组合一起
    """
    def __init__(self, in_dim, n_hidden_1, n_hidden_2, out_dim):
        super(Net, self).__init__()
        self.layer1 = nn.Sequential(nn.Linear(in_dim, n_hidden_1),nn.BatchNorm1d(n_hidden_1))
        self.layer2 = nn.Sequential(nn.Linear(n_hidden_1, n_hidden_2),nn.BatchNorm1d(n_hidden_2))
        self.layer3 = nn.Sequential(nn.Linear(n_hidden_2, out_dim))

    def forward(self, x):
        x = F.relu(self.layer1(x))
        x = F.relu(self.layer2(x))
        x = self.layer3(x)
        return x

## (2)实例化网络
# 检测是否有GPU，有就用，没有就用CPU
device = torch.device("cuda:0" if torch.cuda.if_available else "cpu")
# 实例化网络
model = Net(28*28, 300, 100, 10)
model.to(device)
# 定义损失函数和优化器
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=lr, momentum=momentum)

2.5 训练模型

## 训练模型
# 开始训练
losses = []
acces = []
eval_losses = []
eval_acces = []
print("开始循环，请耐心等待.....")
for epoch in range(num_epoches):
    train_loss = 0
    train_acc = 0
    model.train()
    # 动态修改参数学习率
    if epoch%5==0:
        optimizer.param_groups[0]['lr']*=0.1
    for img, label in train_loader:
        img=img.to(device)
        label = label.to(device)
        img = img.view(img.size(0), -1)
        # 向前传播
        out = model(img)
        loss = criterion(out, label)
        # 反向传播
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        # 记录误差
        train_loss += loss.item()
        # 计算分类的准确率
        _, pred = out.max(1)
        num_correct = (pred == label).sum().item()
        acc = num_correct / img.shape[0]
        train_acc +=acc

    print("第一个循环结束，继续耐心等待....")

    losses.append(train_loss / len(train_loader))
    acces.append(train_acc / len(train_loader))
    # 在测试集上检验效果
    eval_loss = 0
    eval_acc = 0
    # 将模型改为预测模式
    model.eval()
    for img, label in test_loader:
        img=img.to(device)
        label=label.to(device)
        img=img.view(img.size(0),-1)
        out = model(img)
        loss = criterion(out, label)
        # 记录误差
        eval_loss += loss.item()
        # 记录准确率
        _, pred = out.max(1)
        num_correct = (pred == label).sum().item()
        acc = num_correct / img.shape[0]
        eval_acc +=acc

    print("第二个循环结束，准备结束")
    eval_losses.append(eval_loss / len(test_loader))
    eval_acces.append(eval_acc / len(test_loader))
    print('epoch: {}, Train Loss: {:.4f}, Train Acc: {:.4f}, Test Loss: {:.4f}, Test Acc: {:.4f}'.format(epoch, train_loss / len(train_loader), train_acc / len(train_loader), eval_loss / len(test_loader), eval_acc / len(test_loader)))

## 可视化训练结果
plt.title('trainloss')
plt.plot(np.arange(len(losses)), losses)
plt.legend(['Train Loss'], loc='upper right')
print("开始循环，请耐心等待.....")

3 全连接神经网络进行MNIST识别

3.1 数据

import numpy as np
import torch
from torchvision.datasets import mnist
from torch import nn
from torch.autograd import Variable

def data_tf(x):
    x = np.array(x, dtype="float32")/255
    x = (x-0.5)/0.5
    x = x.reshape((-1))   # 这里是为了变为1行，然后m列
    x = torch.from_numpy(x)
    return x

# 下载数据集，有的话就不下载了
train_set = mnist.MNIST("./data",train=True, transform=data_tf, download=False)
test_set = mnist.MNIST("./data",train=False, transform=data_tf, download=False)
a, a_label = train_set[0]
print(a.shape)
print(a_label)

3.2 可视化数据

import matplotlib.pyplot as plt
for i in range(1, 37):
    plt.subplot(6,6,i)
    plt.xticks([])  # 不显示坐标系
    plt.yticks([])
    plt.imshow(train_set.data[i].numpy(), cmap="gray")
    plt.title("%i" % train_set.targets[i])
plt.subplots_adjust(wspace = 0 , hspace = 1)  # 调整
plt.show()

       

from torch.utils.data import DataLoader
train_data = DataLoader(train_set, batch_size=64, shuffle= True)
test_data = DataLoader(test_set, batch_size=128, shuffle=False)
a, a_label = next(iter(train_data))
print(a.shape)
print(a_label.shape)

3.3 定义神经网络

net = nn.Sequential(    nn.Linear(784, 400),
    nn.ReLU(),
    nn.Linear(400, 200),
    nn.ReLU(),
    nn.Linear(200, 100),
    nn.ReLU(),
    nn.Linear(100,10),
    nn.ReLU()
)

if torch.cuda.is_available():
    net = net.cuda()

criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(net.parameters(), 1e-1)

3.4 训练

losses = []
acces  = []
eval_losses = []
eval_acces  = []
# 一共训练20次
for e in range(20):
    train_loss = 0
    train_acc = 0
    net.train()
    for im, label in train_data:
        if torch.cuda.is_available():
            im = Variable(im).cuda()
            label = Variable(label).cuda()
        else:
            im = Variable(im)
            label =Variable(label)

        # 前向传播
        out = net(im)
        loss = criterion(out, label)

        # 反向传播
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        # 误差
        train_loss += loss.item()

        #计算分类的准确率
        # max函数参数1表示按行取最大值，第一个返回值是值，第二个返回值是下标
        # pred是一个固定1*64的向量
        _,pred = out.max(1)
        num_correct = (pred==label).sum().item()
        acc = num_correct/im.shape[0]
        train_acc += acc

    # 此时一轮训练以及完了
    losses.append(train_loss/len(train_data))
    acces.append(train_acc/len(train_data))

    # 在测试集上检验效果
    eval_loss = 0
    eval_acc = 0
    net.eval()
    for im, label in test_data:
        if torch.cuda.is_available():
            im = Variable(im).cuda()
            label = Variable(label).cuda()
        else:
            im = Variable(im)
            label =Variable(label)

        # 前向传播
        out = net(im)
        # 计算误差
        loss = criterion(out, label)
        eval_loss += loss.item()

        # 计算准确率
        _,pred = out.max(1)
        num_correct = (pred==label).sum().item()
        acc = num_correct/im.shape[0]
        eval_acc += acc

    eval_losses.append(eval_loss/len(test_data))
    eval_acces.append(eval_acc/len(test_data))

    print('epoch: {}, Train Loss: {:.6f}, Train Acc: {:.6f}, Eval Loss: {:.6f}, Eval Acc: {:.6f}'.format(e, train_loss / len(train_data), train_acc / len(train_data), eval_loss / len(test_data), eval_acc / len(test_data)))

3.5 展示

%matplotlib inline
plt.subplot(2, 2, 1)
plt.title("train loss")
plt.plot(np.arange(len(losses)), losses)
plt.grid()
plt.subplot(2, 2, 2)
plt.title("train acc")
plt.plot(np.arange(len(acces)), acces)
plt.grid()
plt.subplot(2, 2, 3)
plt.title("test loss")
plt.plot(np.arange(len(eval_losses)), eval_losses)
plt.grid()
plt.subplot(2, 2, 4)
plt.title("test acc")
plt.plot(np.arange(len(eval_acces)), eval_acces)
plt.grid()
plt.subplots_adjust(wspace =0.5, hspace =0.5)

       

for i in range(1, 5):
    im = test_set.data[i]
    label = test_set.targets[i]
    plt.subplot(2, 2, i)
    plt.imshow(im.numpy(), cmap="gray")
    plt.xticks([])
    plt.yticks([])
    im = data_tf(im)
    im = Variable(im).cuda()
    out = net(im)
    _, pred = out.max(0)
    plt.title("outcome=%i" % pred.item())
plt.show()

       

4 如何构建神经网络？

搭建神经网络主要包含：选择网络层，构建网络，选择损失和优化器。

4.1 构建网络层

• 可以在Sequential基础上通过add_module()添加每一层，并且为每一层增加一个单独的名字

• 通过字典的形式添加每一层，设置单独的层名称

class Net(torch.nn.Module):
    def __init__(self):
        super(Net4, self).__init__()
        self.conv = torch.nn.Sequential(
            OrderedDict(
                [
                    ("conv1", torch.nn.Conv2d(3, 32, 3, 1, 1)),
                    ("relu1", torch.nn.ReLU()),
                    ("pool", torch.nn.MaxPool2d(2))
                ]
            ))

        self.dense = torch.nn.Sequential(
            orderedDict([
            ("dense1", torch.nn.Linear(32*3*3,128)),
            ("relu2", torch.nn.ReLU()),
            ("dense2", torch.nn.Linear(128,10))
            ])
        )

4.2 前向、反向传播

4.3 训练模型

5 神经网络工具箱nn

5.2 nn.functional

6 优化器

import torch.optim as optim
optimizer = optimSGD(model.parameters(), lr=lr, momentum=momentum)

out = model(img)
loss = criterion(out, label)

opyimizer.zero_grad()

loss.backward()

optimizer.step()

7 动态修改学习率参数

• 长度1的list
• optimizer.param_groups[0]长度为6的字典，包括权重、lr、momentum等

for epoch in range(num_epoches):
    ## 动态修改参数学习率
    if epoch%5==0
        optimizer.param_groups[0]['lr']*=0.1
        print(optimizer.param_groups[0]['lr'])
    for img, label in train_loader:

8 优化器的比较

## （1）导入需要的模块
import torch
import torch.utils.data as Data
import torch.nn.functional as F
import matplotlib.pyplot as plt
%matplotlib inline
#超参数
LR = 0.01
BATCH_SIZE =32
EPOCH =12
## (2)生成数据
# 生成训练数据
# torch.unsqueeze()作用是将一维变二维，torch只能处理二维数据
x = torch.unsqueeze(torch.linspace(-1, 1, 1000), dim=1)
# 0.1 * torch.normal(x.size()) 增加噪声
y = x.pow(2) + 0.1 * torch.normal(torch.zeros(*x.size()))
torch_dataset = Data.TensorDataset(x,y)
# 一个代批量的生成器
loader = Data.DataLoader(dataset=torch_dataset, batch_size=BATCH_SIZE, shuffle=True)
## (3)构建神经网络
class Net(torch.nn.Module):
    # 初始化
    def __init__(self):
        super(Net, self).__init__()
        self.hidden = torch.nn.Linear(1, 20)
        self.predict = torch.nn.Linear(20, 1)

    # 向前传递
    def forward(self, x):
        x = F.relu(self.hidden(x))
        x = self.predict(x)
        return x

## (4)使用多种优化器
net_SGD = Net()
net_Momentum = Net()
net_RMSProp = Net()
net_Adam = Net()
nets = [net_SGD, net_Momentum, net_RMSProp, net_Adam]
opt_SGD =torch.optim.SGD(net_SGD.parameters(), lr=LR)
opt_Momentum =torch.optim.SGD(net_Momentum.parameters(), lr=LR, momentum = 0.9)
opt_RMSProp =torch.optim.RMSprop(net_RMSProp.parameters(), lr=LR, alpha = 0.9)
opt_Adam =torch.optim.Adam(net_Adam.parameters(), lr=LR, betas=(0.9, 0.99))
optimizers = [opt_SGD, opt_Momentum, opt_RMSProp, opt_Adam]

## (5)训练模型
loss_func = torch.nn.MSELoss()
loss_his = [[], [], [], []]
for epoch in range(EPOCH):
    for step, (batch_x, batch_y) in enumerate(loader):
        for net, opt, l_his in zip(nets, optimizers, loss_his):
            output = net(batch_x)
            loss = loss_func(output, batch_y)
            opt.zero_grad()
            loss.backward()
            opt.step()
            l_his.append(loss.data.numpy())
labels = ['SGD', 'Momentum', 'RMSProp', 'Adam']
## (6)可视化结果
for i, l_his in enumerate(loss_his):
    plt.plot(l_his, label=labels[i])
plt.legend(loc='best')
plt.xlabel('Steps')
plt.ylabel('Loss')
plt.ylim((0, 0.2))
plt.show()

       

推荐阅读

• 基于Pytorch的动态卷积复现
  

• 当代研究生应当掌握的5种Pytorch并行训练方法（单机多卡）
  

• 综述｜核心开发者全面解读Pytorch内部机制
  

ACCV 2020国际细粒度网络图像识别竞赛即将开赛！

添加极市小助手微信（ID : cvmart2），备注：姓名-学校/公司-研究方向-城市（如：小极-北大-目标检测-深圳），即可申请加入极市目标检测/图像分割/工业检测/人脸/医学影像/3D/SLAM/自动驾驶/超分辨率/姿态估计/ReID/GAN/图像增强/OCR/视频理解等技术交流群：每月大咖直播分享、真实项目需求对接、求职内推、算法竞赛、干货资讯汇总、与 10000+来自港科大、北大、清华、中科院、CMU、腾讯、百度等名校名企视觉开发者互动交流~

△长按添加极市小助手

△长按关注极市平台，获取最新CV干货

觉得有用麻烦给个在看啦~