Datawhale 组对学习打卡营 任务3：多层感知机

模型

激活函数

1. RuLe
2. Sigmoid
3. tanh

pytorch实现多层感知机

import torch
from torch import nn
from torch.nn import init
import numpy as np
import sys
import d2lzh1981 as d2l    # Data团队提供的包

# *****初始化模型和各个参数*****
num_inputs, num_outputs, num_hiddens = 784, 10, 256
    
nets = nn.Sequential(
        d2l.FlattenLayer(), # 特征的转换
        nn.Linear(num_inputs, num_hiddens), #输入层到隐藏层的线性回归模型
        nn.ReLU(),  ## 定义激活函数
        nn.Linear(num_hiddens, num_outputs), # 隐藏层到输出层的线性回归模型
        )
    
# 给网络初始化参数
for params in nets.parameters():
    init.normal_(params, mean=0, std=0.01)

# *****训练*****
batch_size = 256
# 获取数据
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size,root='/home/kesci/input/FashionMNIST2065')
# 定义损失函数
loss = torch.nn.CrossEntropyLoss()

# 定义优化函数，选用梯度下降
optimizer = torch.optim.SGD(nets.parameters(), lr=0.5)

# 定义训练周期
num_epochs = 5
d2l.train_ch3(nets, train_iter, test_iter, loss, num_epochs, batch_size, None, None, optimizer)