《动手学深度学习》 NIN with gluon

NIN

Net in Net利用1x1的Conv来代替Dense，Conv的输入输出维数是4维，Dense的输入输出维数是2维，这样就避免了在Dense后接Conv时还需要将输出的维数从2维改至4维，数据量过大，导致计算负荷过重。最重要的是提出了一种思想。

代码

import d2lzh as d2l
from mxnet import gluon, init, nd, autograd
from mxnet.gluon import nn
import time


def nin_block(num_channels, kernel_size, padding, strides=1, max_pooling=True):
    blk = nn.Sequential()
    blk.add(nn.Conv2D(num_channels, kernel_size,
                      strides, padding, activation='relu'),
            nn.Conv2D(num_channels, kernel_size=1, activation='relu'),
            nn.Conv2D(num_channels, kernel_size=1, activation='relu'))
    if max_pooling:
        blk.add(nn.MaxPool2D(pool_size=3, strides=2))
    return blk


net = nn.Sequential()
net.add(nin_block(96, 11, 0, 4, True),
        nin_block(256, 5, 2, 1, True),
        nin_block(384, 3, 1, 1, True),
        nn.Dropout(0.5),
        nin_block(10, 3, 1, 1, False),
        nn.GlobalAvgPool2D(),
        nn.Flatten())

batch_size = 64
# 构建数据集，将原来28x28的图片放大到224x224
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size, resize=224)

ctx = d2l.try_gpu()
net.initialize(ctx=ctx, init=init.Xavier())
print('training on', ctx)

softmax_cross_entropy = gluon.loss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(), 'sgd', {'learning_rate': 0.1})

for epoch in range(3):
    train_loss_sum = 0
    train_acc_sum = 0
    n = 0
    start = time.time()
    for X, y in train_iter:
        X, y = X.as_in_context(ctx), y.as_in_context(ctx)
        with autograd.record():
            y_hat = net(X)
            loss = softmax_cross_entropy(y_hat, y).sum()
        loss.backward()
        trainer.step(batch_size)
        y = y.astype('float32')
        train_loss_sum += loss.asscalar()
        train_acc_sum += (y_hat.argmax(axis=1) == y).sum().asscalar()
        n += y.size
    test_acc = d2l.evaluate_accuracy(test_iter, net, ctx)
    print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f, time %.1f sec'
          % (epoch + 1, train_loss_sum / n, train_acc_sum / n, test_acc, time.time() - start))

结果

training on gpu(0)
epoch 1, loss 1.9699, train acc 0.294, test acc 0.469, time 130.4 sec
epoch 2, loss 1.0835, train acc 0.613, test acc 0.736, time 126.1 sec
epoch 3, loss 0.6597, train acc 0.749, test acc 0.790, time 127.1 sec