SoftMax回归（李沐老师课程）

SoftMax回归

分类的数据集

分类和回归的区别

从回归到多类分类-均方损失

softmax损失函数推导

损失函数

图片分类数据集Fashion-MNIST

import torch
import torchvision
from torch.utils import data
from torchvision import transforms
from d2l import torch as d2l
​
d2l.use_svg_display()
​
"""
通过框架中的内置函数将Fashion-MNIST数据集进行下载
通过ToTrnsor实例将图像数据从PIL类型变换成为32浮点数据格式
并除以255使得所有像素点位于0-1
"""
trans = transforms.ToTensor()
mnist_train = torchvision.datasets.FashionMNIST(root='./data', train=True, transform=trans, download=True)
miist_test = torchvision.datasets.FashionMNIST(root='./data', train=False, transform=trans, download=True)
print(len(mnist_train), len(mnist_train))
print(mnist_train[0][0].shape)
​
​
def get_fashion_mnist_labels(labels):
    """
    返回Fashion-MNIST数据集的文本标签
    :param labels:
    :return:
    """
    text_labels = ['t-shirt', 'trousers', 'pullover', 'dress', 'coat', 'sandals', 'shirt', 'sneaker', 'bag',
                   'ankle boot']
    return [text_labels[int(i)] for i in labels]
​
​
def show_images(imgs, num_rows, num_cols, title=None, scale=1.5):
    """
    Plot a list of images.
    :param img:
    :param num_rows:
    :param num_cols:
    :param title:
    :param scale:
    :return:
    """
    figsize = (num_cols * scale, num_rows * scale)
    _, axes = d2l.plt.subplots(num_rows, num_cols, figsize=figsize)
    axes = axes.flatten()
    for i, (ax, img) in enumerate(zip(axes, imgs)):
        if torch.is_tensor(img):
            # 图片张量
            ax.imshow(img.numpy())
        else:
            ax.imshow(img)
​
​
X, y = next(iter(data.DataLoader(mnist_train, batch_size=18)))
show_images(X.reshape(18, 28, 28), 2, 9, title=get_fashion_mnist_labels(y))
d2l.plt.show()
batch_size = 256
​
​
def get_dataloader_workers():
    """使用四个进程来读取数据"""
    return 4
​
​
train_iter = data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True, num_workers=get_dataloader_workers())
timer = d2l.Timer()
for X, y in train_iter:
    continue
print(f'{timer.stop():.2f} sec')
​
​
def load_data_fashion_mnist(batch_size, resize=None):
    """下载Fashion-MNIST数据集，然后将其加载到内存中"""
    trans = [transforms.ToTensor()]
    if resize:
        trans.insert(0, transforms.Resize(resize))
    trans = transforms.Compose(trans)
    mnist_train = torchvision.datasets.FashionMNIST(root='./data', train=True, transform=trans, download=True)
    mnist_test = torchvision.datasets.FashionMNIST(root='./data', train=False, transform=trans, download=True)
    return ((data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True, num_workers=get_dataloader_workers())),data.DataLoader(mnist_test, batch_size=batch_size, shuffle=True, num_workers=get_dataloader_workers()))