3.10. 多层感知机的简洁实现

import d2lzh as d2l
from mxnet import gluon, init

net = nn.Sequential()
net.add(nn.Dense(256, activation='relu'),
        nn.Dense(10))
net.initialize(init.Normal(sigma=0.01))

batch_size = 256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
loss = gloss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(), 'sgd', {'learning_rate': 0.5})
num_epochs = 5
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, batch_size, None,
              None, trainer)

• 通过Gluon可以更简洁地实现多层感知机。

• 使用其他的激活函数，看看对结果的影响。