Pytorch【60天修煉計劃】之第二階段——進階：卷積神經網絡之LeNet

LeNet是來自論文《Gradient-Based Learning Applied to Document Recognition》中提出的網絡，它是CNN的開山鼻祖，對於手寫數字的識別有了當時最先進的結果。

這是LeNet的網絡結構：

• input [ 1, 32 × 32 ] ( channels, weight * height)
• Conv2d [ 1, 6, 5 × 5 ] (in_channels, out_channels, kernel_size)
• Maxpool2d [ 6, 2 × 2, 2 ] (channels, kernel_size, stride)
• Conv2d [ 6, 16, 5 × 5 ]
• Maxpool2d [ 16, 2 × 2, 2 ]
• FC [ 16 × 4 × 4, 120 ]
• FC [ 120, 84 ]
• FC [ 84, 10 ]

項目結構圖

模型

按照上面對於網絡結構的分解，在pytorch構建如下網絡：

class LeNet(nn.Module):
    def __init__(self):
        super(LeNet, self).__init__()
        self.conv = nn.Sequential(
            nn.Conv2d(1, 6, 5),
            nn.Sigmoid(),
            nn.MaxPool2d(2, 2),
            nn.Conv2d(6, 16, 5),
            nn.Sigmoid(),
            nn.MaxPool2d(2, 2)
        )
        self.fc = nn.Sequential(
            nn.Linear(16*4*4, 120),
            nn.Sigmoid(),
            nn.Linear(120, 84),
            nn.Sigmoid(),
            nn.Linear(84, 10)
        )

    def forward(self ,X):
        feature = self.conv(X)
        output = self.fc(feature.view(X.shape[0], -1))
        return output

訓練數據

def train(net, train_iter, test_iter, batch_size, optimizer, device, num_epochs):
    net = net.to(device)
    print("training on ", device)
    loss = nn.CrossEntropyLoss()
    batch_count = 0
    for epoch in range(num_epochs):
        train_l_sum, train_acc_sum, n, start = 0.0, 0.0, 0, time.time()
        for X, y in train_iter:
            # 放入設備中
            X = X.to(device)
            y = y.to(device)
            # 由模型得到的值
            y_hat = net(X)
            # 求損失
            l = loss(y_hat, y)
            optimizer.zero_grad()
            l.backward()
            optimizer.step()
            train_l_sum += l.cpu().item()
            train_acc_sum += (y_hat.argmax(dim=1) == y).sum().cpu().item()
            n += y.shape[0]
            batch_count += 1

        test_acc = evaluate_accuracy(test_iter, net)
        print(train_l_sum, train_acc_sum)
        print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f,time % .1f sec' % (epoch + 1, train_l_sum / batch_count, train_acc_sum / n, test_acc, time.time() - start))