{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"name": "Batch Normalization Evaluation (with Residual Connection)",
"provenance": [],
"collapsed_sections": [],
"authorship_tag": "ABX9TyO1gf/cdVb/RgD6vps+WTBN",
"include_colab_link": true
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
},
"accelerator": "GPU"
},
"cells": [
{
"cell_type": "markdown",
"metadata": {
"id": "view-in-github",
"colab_type": "text"
},
"source": [
"
"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "K3i3vfP0Duid"
},
"source": [
"#### ResNet18 모델 정의 및 인스턴스 초기화\r\n",
"\r\n",
"* 일반적인 [ResNet18](https://arxiv.org/abs/1512.03385) 모델 아키텍처를 사용합니다.\r\n",
"* 배치 정규화(batch normalization) 사용 여부를 설정할 수 있습니다.\r\n",
"* [Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift](https://arxiv.org/abs/1502.03167)"
]
},
{
"cell_type": "code",
"metadata": {
"id": "HyGPurUyDwZs"
},
"source": [
"import torch\r\n",
"import torch.nn as nn\r\n",
"import torch.nn.functional as F\r\n",
"import torch.backends.cudnn as cudnn\r\n",
"import torch.optim as optim\r\n",
"import os\r\n",
"import matplotlib.pyplot as plt\r\n",
"\r\n",
"\r\n",
"# ResNet18을 위해 최대한 간단히 수정한 BasicBlock 클래스 정의\r\n",
"class BasicBlock(nn.Module):\r\n",
" def __init__(self, in_planes, planes, with_BN, stride=1):\r\n",
" super(BasicBlock, self).__init__()\r\n",
" self.with_BN = with_BN\r\n",
"\r\n",
" # 3x3 필터를 사용 (너비와 높이를 줄일 때는 stride 값 조절)\r\n",
" self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)\r\n",
" if with_BN: # 배치 정규화(batch normalization)를 사용하는 경우\r\n",
" self.bn1 = nn.BatchNorm2d(planes) \r\n",
"\r\n",
" # 3x3 필터를 사용 (패딩을 1만큼 주기 때문에 너비와 높이가 동일)\r\n",
" self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)\r\n",
" if with_BN: # 배치 정규화(batch normalization)를 사용하는 경우\r\n",
" self.bn2 = nn.BatchNorm2d(planes)\r\n",
"\r\n",
" self.shortcut = nn.Sequential() # 단순한 identity mapping인 경우\r\n",
" if stride != 1: # stride가 1이 아니라면, Identity mapping이 아닌 경우\r\n",
" modules = [nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride, bias=False)]\r\n",
" if with_BN:\r\n",
" modules.append(nn.BatchNorm2d(planes))\r\n",
" self.shortcut = nn.Sequential(*modules)\r\n",
"\r\n",
" def forward(self, x):\r\n",
" if self.with_BN: # 배치 정규화(batch normalization)를 사용하는 경우\r\n",
" out = F.relu(self.bn1(self.conv1(x)))\r\n",
" out = self.bn2(self.conv2(out))\r\n",
" else:\r\n",
" out = F.relu(self.conv1(x))\r\n",
" out = self.conv2(out)\r\n",
" out += self.shortcut(x) # (핵심) skip connection\r\n",
" out = F.relu(out)\r\n",
" return out\r\n",
"\r\n",
"\r\n",
"# ResNet 클래스 정의\r\n",
"class ResNet(nn.Module):\r\n",
" def __init__(self, block, num_blocks, with_BN, num_classes=10):\r\n",
" super(ResNet, self).__init__()\r\n",
" self.in_planes = 64\r\n",
" self.with_BN = with_BN\r\n",
"\r\n",
" # 64개의 3x3 필터(filter)를 사용\r\n",
" self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False)\r\n",
" if with_BN: # 배치 정규화(batch normalization)를 사용하는 경우\r\n",
" self.bn1 = nn.BatchNorm2d(64)\r\n",
" self.layer1 = self._make_layer(block, 64, with_BN, num_blocks[0], stride=1)\r\n",
" self.layer2 = self._make_layer(block, 128, with_BN, num_blocks[1], stride=2)\r\n",
" self.layer3 = self._make_layer(block, 256, with_BN, num_blocks[2], stride=2)\r\n",
" self.layer4 = self._make_layer(block, 512, with_BN, num_blocks[3], stride=2)\r\n",
" self.linear = nn.Linear(512, num_classes)\r\n",
"\r\n",
" def _make_layer(self, block, planes, with_BN, num_blocks, stride):\r\n",
" strides = [stride] + [1] * (num_blocks - 1)\r\n",
" layers = []\r\n",
" for stride in strides:\r\n",
" layers.append(block(self.in_planes, planes, with_BN, stride))\r\n",
" self.in_planes = planes # 다음 레이어를 위해 채널 수 변경\r\n",
" return nn.Sequential(*layers)\r\n",
"\r\n",
" def forward(self, x):\r\n",
" if self.with_BN: # 배치 정규화(batch normalization)를 사용하는 경우\r\n",
" out = F.relu(self.bn1(self.conv1(x)))\r\n",
" else:\r\n",
" out = F.relu(self.conv1(x))\r\n",
" out = self.layer1(out)\r\n",
" out = self.layer2(out)\r\n",
" out = self.layer3(out)\r\n",
" out = self.layer4(out)\r\n",
" out = F.avg_pool2d(out, 4)\r\n",
" out = out.view(out.size(0), -1)\r\n",
" out = self.linear(out)\r\n",
" return out\r\n",
"\r\n",
"\r\n",
"# ResNet18 함수 정의\r\n",
"def ResNet18(with_BN):\r\n",
" return ResNet(BasicBlock, [2, 2, 2, 2], with_BN)"
],
"execution_count": 1,
"outputs": []
},
{
"cell_type": "markdown",
"metadata": {
"id": "rjC1ZQ6RIfcL"
},
"source": [
"#### 데이터셋(Dataset) 다운로드 및 불러오기\r\n",
"\r\n",
"* 실험을 위해 CIFAR-10 데이터셋을 사용합니다."
]
},
{
"cell_type": "code",
"metadata": {
"id": "-KEhEyohIkKs",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "08c6247b-1f0e-4020-aeea-4af530b4cca6"
},
"source": [
"import torchvision\r\n",
"import torchvision.transforms as transforms\r\n",
"\r\n",
"transform_train = transforms.Compose([\r\n",
" transforms.RandomCrop(32, padding=4),\r\n",
" transforms.RandomHorizontalFlip(),\r\n",
" transforms.ToTensor(),\r\n",
"])\r\n",
"\r\n",
"transform_test = transforms.Compose([\r\n",
" transforms.ToTensor(),\r\n",
"])\r\n",
"\r\n",
"train_dataset = torchvision.datasets.CIFAR10(root='./data', train=True, download=True, transform=transform_train)\r\n",
"test_dataset = torchvision.datasets.CIFAR10(root='./data', train=False, download=True, transform=transform_test)\r\n",
"\r\n",
"train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=128, shuffle=True, num_workers=4)\r\n",
"test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=100, shuffle=False, num_workers=4)"
],
"execution_count": 3,
"outputs": [
{
"output_type": "stream",
"text": [
"Files already downloaded and verified\n",
"Files already downloaded and verified\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "GXcWfBZyIm6p"
},
"source": [
"#### 학습(Training) 및 테스트(Test) 함수 정의"
]
},
{
"cell_type": "code",
"metadata": {
"id": "vx7Ruf7LHWH9",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "772f1ce2-6198-43c5-e411-f2e9db30b79d"
},
"source": [
"criterion = nn.CrossEntropyLoss() # 분류(classification) 문제\r\n",
"\r\n",
"\r\n",
"def train(net, optimizer, epoch, step):\r\n",
" net.train()\r\n",
" correct = 0 # 정답을 맞힌 이미지 개수\r\n",
" total = 0 # 전체 이미지 개수\r\n",
" steps = [] # 학습 스텝(step)\r\n",
" losses = [] # 각 스텝에서의 손실(loss)\r\n",
"\r\n",
" for _, (inputs, targets) in enumerate(train_loader):\r\n",
" inputs, targets = inputs.cuda(), targets.cuda()\r\n",
"\r\n",
" optimizer.zero_grad()\r\n",
" outputs = net(inputs)\r\n",
" loss = criterion(outputs, targets)\r\n",
" loss.backward()\r\n",
" optimizer.step()\r\n",
" \r\n",
" _, predicted = outputs.max(1)\r\n",
" correct += predicted.eq(targets).sum().item()\r\n",
" total += targets.size(0)\r\n",
"\r\n",
" steps.append(step)\r\n",
" losses.append(loss.item())\r\n",
" step += 1\r\n",
"\r\n",
" return correct / total, steps, losses\r\n",
"\r\n",
"\r\n",
"def test(net, optimizer, epoch):\r\n",
" net.eval()\r\n",
" correct = 0 # 정답을 맞힌 이미지 개수\r\n",
" total = 0 # 전체 이미지 개수\r\n",
" loss = 0 # 손실(loss)\r\n",
"\r\n",
" for batch_idx, (inputs, targets) in enumerate(test_loader):\r\n",
" inputs, targets = inputs.cuda(), targets.cuda()\r\n",
"\r\n",
" outputs = net(inputs)\r\n",
" loss += criterion(outputs, targets).item()\r\n",
"\r\n",
" _, predicted = outputs.max(1)\r\n",
" correct += predicted.eq(targets).sum().item()\r\n",
" total += targets.size(0)\r\n",
"\r\n",
" return correct / total, loss"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "f-U29olPI9x-"
},
"source": [
"#### Without BN: 학습(Training) 진행\r\n",
"\r\n",
"* 먼저 BatchNorm을 적용하지 않은 모델의 성능을 평가합니다.\r\n",
"* ([참고](https://arxiv.org/abs/1712.09913)) 기본적으로 Residual Connection은 Loss Landscape Smoothing 효과가 있다는 점을 감안할 필요가 있습니다."
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "fFSqQWYiUnaw",
"outputId": "21d081e0-110d-424a-aa5e-140aac9e237c"
},
"source": [
"print('모델 파라미터를 초기화합니다.')\r\n",
"net = ResNet18(with_BN=False).cuda()\r\n",
"learning_rate = 0.01\r\n",
"optimizer = optim.SGD(net.parameters(), lr=learning_rate, momentum=0.9, weight_decay=0.0002)\r\n",
"\r\n",
"total_params = sum(p.numel() for p in net.parameters() if p.requires_grad)\r\n",
"print('학습 가능한 총 파라미터 수:', total_params)"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"모델 파라미터를 초기화합니다.\n",
"학습 가능한 총 파라미터 수: 11164362\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "BZ2BXz0pIAjz",
"outputId": "03fbe68d-1098-4691-fac1-a5a78aa1479d"
},
"source": [
"without_BN_train_accuracies = []\r\n",
"without_BN_steps = []\r\n",
"without_BN_train_losses = []\r\n",
"without_BN_test_accuracies = []\r\n",
"without_BN_test_losses = []\r\n",
"\r\n",
"epochs = 20\r\n",
"\r\n",
"for epoch in range(0, epochs):\r\n",
" print(f'[ Epoch: {epoch}/{epochs} ]')\r\n",
" train_accuracy, steps, train_losses = train(net, optimizer, epoch, len(without_BN_steps))\r\n",
" without_BN_train_accuracies.append(train_accuracy)\r\n",
" without_BN_steps.extend(steps)\r\n",
" without_BN_train_losses.extend(train_losses)\r\n",
" print(f'Train accuracy = {train_accuracy * 100:.2f} / Train loss = {sum(train_losses)}')\r\n",
"\r\n",
" test_accuracy, test_loss = test(net, optimizer, epoch)\r\n",
" without_BN_test_accuracies.append(test_accuracy)\r\n",
" without_BN_test_losses.append(test_loss)\r\n",
" print(f'Test accuracy = {test_accuracy * 100:.2f} / Test loss = {test_loss}')"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"[ Epoch: 0/20 ]\n",
"Train accuracy = 16.20 / Train loss = 857.8076474666595\n",
"Test accuracy = 19.48 / Test loss = 213.57228648662567\n",
"[ Epoch: 1/20 ]\n",
"Train accuracy = 27.93 / Train loss = 759.2249348163605\n",
"Test accuracy = 31.69 / Test loss = 184.4820580482483\n",
"[ Epoch: 2/20 ]\n",
"Train accuracy = 34.92 / Train loss = 692.4940202236176\n",
"Test accuracy = 39.16 / Test loss = 160.85358726978302\n",
"[ Epoch: 3/20 ]\n",
"Train accuracy = 40.19 / Train loss = 633.2600127458572\n",
"Test accuracy = 43.34 / Test loss = 156.79514908790588\n",
"[ Epoch: 4/20 ]\n",
"Train accuracy = 43.83 / Train loss = 597.2320219278336\n",
"Test accuracy = 47.34 / Test loss = 145.44127428531647\n",
"[ Epoch: 5/20 ]\n",
"Train accuracy = 47.55 / Train loss = 563.2948558330536\n",
"Test accuracy = 50.86 / Test loss = 135.2007726430893\n",
"[ Epoch: 6/20 ]\n",
"Train accuracy = 52.13 / Train loss = 519.5157498121262\n",
"Test accuracy = 55.02 / Test loss = 123.66149139404297\n",
"[ Epoch: 7/20 ]\n",
"Train accuracy = 56.03 / Train loss = 481.54733526706696\n",
"Test accuracy = 58.77 / Test loss = 115.929270029068\n",
"[ Epoch: 8/20 ]\n",
"Train accuracy = 59.31 / Train loss = 449.87207156419754\n",
"Test accuracy = 60.58 / Test loss = 110.74408739805222\n",
"[ Epoch: 9/20 ]\n",
"Train accuracy = 62.34 / Train loss = 416.18682181835175\n",
"Test accuracy = 63.75 / Test loss = 100.98181921243668\n",
"[ Epoch: 10/20 ]\n",
"Train accuracy = 64.72 / Train loss = 393.7053857445717\n",
"Test accuracy = 64.46 / Test loss = 101.189976811409\n",
"[ Epoch: 11/20 ]\n",
"Train accuracy = 67.36 / Train loss = 365.18211060762405\n",
"Test accuracy = 66.79 / Test loss = 94.47453755140305\n",
"[ Epoch: 12/20 ]\n",
"Train accuracy = 69.20 / Train loss = 343.7658784389496\n",
"Test accuracy = 70.43 / Test loss = 83.89928996562958\n",
"[ Epoch: 13/20 ]\n",
"Train accuracy = 71.07 / Train loss = 325.12918388843536\n",
"Test accuracy = 71.35 / Test loss = 82.74367016553879\n",
"[ Epoch: 14/20 ]\n",
"Train accuracy = 72.70 / Train loss = 308.0616769194603\n",
"Test accuracy = 74.11 / Test loss = 75.04425463080406\n",
"[ Epoch: 15/20 ]\n",
"Train accuracy = 74.02 / Train loss = 290.66746068000793\n",
"Test accuracy = 73.42 / Test loss = 78.14051738381386\n",
"[ Epoch: 16/20 ]\n",
"Train accuracy = 75.71 / Train loss = 275.05585649609566\n",
"Test accuracy = 75.07 / Test loss = 73.68815511465073\n",
"[ Epoch: 17/20 ]\n",
"Train accuracy = 76.89 / Train loss = 262.2904309928417\n",
"Test accuracy = 75.23 / Test loss = 73.28887414932251\n",
"[ Epoch: 18/20 ]\n",
"Train accuracy = 77.84 / Train loss = 248.9377856850624\n",
"Test accuracy = 77.07 / Test loss = 66.20125043392181\n",
"[ Epoch: 19/20 ]\n",
"Train accuracy = 79.22 / Train loss = 237.0634382069111\n",
"Test accuracy = 78.53 / Test loss = 64.17613476514816\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "AeXK5wgydIxn"
},
"source": [
"* 학습 과정에서의 스텝(step)에 따른 손실(loss) 값을 시각화합니다."
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 295
},
"id": "pgNM6qIhcMbl",
"outputId": "29cdf2dc-1ebb-408a-ce06-48ade025031d"
},
"source": [
"plt.plot(without_BN_steps, without_BN_train_losses)\r\n",
"plt.title('Train Loss')\r\n",
"plt.xlabel('Step')\r\n",
"plt.ylabel('Loss')\r\n",
"plt.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "4suCN4-SgX9n"
},
"source": [
"* 테스트 정확도(accuracy) 및 손실(loss) 값을 시각화합니다."
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 573
},
"id": "mWZ-AOHlg7Vz",
"outputId": "4d491b55-81e3-48b1-e424-21e95a95ba4b"
},
"source": [
"plt.plot([i for i in range(len(without_BN_test_accuracies))], without_BN_test_accuracies)\r\n",
"plt.title('Test Accuracy')\r\n",
"plt.xlabel('Epoch')\r\n",
"plt.ylabel('Accuracy')\r\n",
"plt.show()\r\n",
"\r\n",
"plt.plot([i for i in range(len(without_BN_test_losses))], without_BN_test_losses)\r\n",
"plt.title('Test Loss')\r\n",
"plt.xlabel('Epoch')\r\n",
"plt.ylabel('Loss')\r\n",
"plt.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
},
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "7JNxL2haUrJL"
},
"source": [
"#### With BN: 학습(Training) 진행\r\n",
"\r\n",
"* 이어서 BatchNorm을 적용했을 때의 성능을 평가합니다."
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "_KcuJXIeUrzP",
"outputId": "e7c5c38f-b4f6-4f38-ff66-58bd69f68a60"
},
"source": [
"print('모델 파라미터를 초기화합니다.')\r\n",
"net = ResNet18(with_BN=True).cuda()\r\n",
"learning_rate = 0.01\r\n",
"optimizer = optim.SGD(net.parameters(), lr=learning_rate, momentum=0.9, weight_decay=0.0002)\r\n",
"\r\n",
"total_params = sum(p.numel() for p in net.parameters() if p.requires_grad)\r\n",
"print('학습 가능한 총 파라미터 수:', total_params)"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"모델 파라미터를 초기화합니다.\n",
"학습 가능한 총 파라미터 수: 11173962\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "iKzavdMWID0z",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "a9581d2c-3a26-40bd-f6b5-45ded5ab2ee5"
},
"source": [
"with_BN_train_accuracies = []\r\n",
"with_BN_steps = []\r\n",
"with_BN_train_losses = []\r\n",
"with_BN_test_accuracies = []\r\n",
"with_BN_test_losses = []\r\n",
"\r\n",
"epochs = 20\r\n",
"\r\n",
"for epoch in range(0, epochs):\r\n",
" print(f'[ Epoch: {epoch}/{epochs} ]')\r\n",
" train_accuracy, steps, train_losses = train(net, optimizer, epoch, len(with_BN_steps))\r\n",
" with_BN_train_accuracies.append(train_accuracy)\r\n",
" with_BN_steps.extend(steps)\r\n",
" with_BN_train_losses.extend(train_losses)\r\n",
" print(f'Train accuracy = {train_accuracy * 100:.2f} / Train loss = {sum(train_losses)}')\r\n",
"\r\n",
" test_accuracy, test_loss = test(net, optimizer, epoch)\r\n",
" with_BN_test_accuracies.append(test_accuracy)\r\n",
" with_BN_test_losses.append(test_loss)\r\n",
" print(f'Test accuracy = {test_accuracy * 100:.2f} / Test loss = {test_loss}')"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"[ Epoch: 0/20 ]\n",
"Train accuracy = 48.11 / Train loss = 557.1330206990242\n",
"Test accuracy = 43.08 / Test loss = 216.10532009601593\n",
"[ Epoch: 1/20 ]\n",
"Train accuracy = 67.55 / Train loss = 358.28609424829483\n",
"Test accuracy = 67.67 / Test loss = 98.73264443874359\n",
"[ Epoch: 2/20 ]\n",
"Train accuracy = 75.23 / Train loss = 278.1966603696346\n",
"Test accuracy = 76.13 / Test loss = 68.25526139140129\n",
"[ Epoch: 3/20 ]\n",
"Train accuracy = 79.19 / Train loss = 235.17466029524803\n",
"Test accuracy = 73.37 / Test loss = 83.11967742443085\n",
"[ Epoch: 4/20 ]\n",
"Train accuracy = 81.87 / Train loss = 202.6331698000431\n",
"Test accuracy = 73.12 / Test loss = 86.56801211833954\n",
"[ Epoch: 5/20 ]\n",
"Train accuracy = 84.06 / Train loss = 179.89355811476707\n",
"Test accuracy = 81.55 / Test loss = 54.92713603377342\n",
"[ Epoch: 6/20 ]\n",
"Train accuracy = 85.43 / Train loss = 166.3792644739151\n",
"Test accuracy = 81.26 / Test loss = 60.015089720487595\n",
"[ Epoch: 7/20 ]\n",
"Train accuracy = 86.72 / Train loss = 149.63439264893532\n",
"Test accuracy = 84.66 / Test loss = 46.31218422949314\n",
"[ Epoch: 8/20 ]\n",
"Train accuracy = 87.95 / Train loss = 135.99417962133884\n",
"Test accuracy = 85.05 / Test loss = 46.0186281055212\n",
"[ Epoch: 9/20 ]\n",
"Train accuracy = 88.79 / Train loss = 127.5717298835516\n",
"Test accuracy = 85.79 / Test loss = 42.48643907904625\n",
"[ Epoch: 10/20 ]\n",
"Train accuracy = 89.58 / Train loss = 118.30104224383831\n",
"Test accuracy = 85.43 / Test loss = 43.56877526640892\n",
"[ Epoch: 11/20 ]\n",
"Train accuracy = 89.96 / Train loss = 111.83856330811977\n",
"Test accuracy = 85.08 / Test loss = 48.1028977483511\n",
"[ Epoch: 12/20 ]\n",
"Train accuracy = 90.89 / Train loss = 102.58311522006989\n",
"Test accuracy = 87.91 / Test loss = 37.35952450335026\n",
"[ Epoch: 13/20 ]\n",
"Train accuracy = 91.33 / Train loss = 95.34802985191345\n",
"Test accuracy = 86.42 / Test loss = 43.105433121323586\n",
"[ Epoch: 14/20 ]\n",
"Train accuracy = 91.95 / Train loss = 89.77752082794905\n",
"Test accuracy = 87.79 / Test loss = 39.972014874219894\n",
"[ Epoch: 15/20 ]\n",
"Train accuracy = 92.40 / Train loss = 85.31641776114702\n",
"Test accuracy = 85.86 / Test loss = 47.14975622296333\n",
"[ Epoch: 16/20 ]\n",
"Train accuracy = 92.96 / Train loss = 79.86298192292452\n",
"Test accuracy = 88.58 / Test loss = 36.66135448217392\n",
"[ Epoch: 17/20 ]\n",
"Train accuracy = 93.06 / Train loss = 76.22297156602144\n",
"Test accuracy = 88.07 / Test loss = 40.42744205892086\n",
"[ Epoch: 18/20 ]\n",
"Train accuracy = 93.56 / Train loss = 72.37405498325825\n",
"Test accuracy = 88.73 / Test loss = 36.78540977835655\n",
"[ Epoch: 19/20 ]\n",
"Train accuracy = 94.19 / Train loss = 66.42253036051989\n",
"Test accuracy = 87.94 / Test loss = 42.450388342142105\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "s3YAOnZQlg8e"
},
"source": [
"* 학습 과정에서의 스텝(step)에 따른 손실(loss) 값을 시각화합니다."
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 295
},
"id": "S_hWBA18lrqY",
"outputId": "b0053c09-653f-4811-9ee3-c96529e786b7"
},
"source": [
"plt.plot(with_BN_steps, with_BN_train_losses)\r\n",
"plt.title('Train Loss')\r\n",
"plt.xlabel('Step')\r\n",
"plt.ylabel('Loss')\r\n",
"plt.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "WwImoDSsl65r"
},
"source": [
"* 테스트 정확도(accuracy) 및 손실(loss) 값을 시각화합니다."
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 573
},
"id": "kskZxBh9lr_3",
"outputId": "3db977f5-ac84-46b8-a967-252880c9493a"
},
"source": [
"plt.plot([i for i in range(len(with_BN_test_accuracies))], with_BN_test_accuracies)\r\n",
"plt.title('Test Accuracy')\r\n",
"plt.xlabel('Epoch')\r\n",
"plt.ylabel('Accuracy')\r\n",
"plt.show()\r\n",
"\r\n",
"plt.plot([i for i in range(len(with_BN_test_losses))], with_BN_test_losses)\r\n",
"plt.title('Test Loss')\r\n",
"plt.xlabel('Epoch')\r\n",
"plt.ylabel('Loss')\r\n",
"plt.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
},
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "yE0qlInxmnYf"
},
"source": [
"#### Without BN과 With BN 비교하기\r\n",
"\r\n",
"* 배치 정규화(batch normalization)를 이용할 때 손실 값의 변동 폭이 작습니다.\r\n",
"* 성능이 훨씬 좋아지는 것을 확인할 수 있으며 초반에도 빠른 성능 향상을 보입니다."
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 295
},
"id": "JP-CL88hmpKK",
"outputId": "2b3de677-af0d-4586-f91a-34c5d68eeab0"
},
"source": [
"plt.plot(without_BN_steps, without_BN_train_losses)\r\n",
"plt.plot(with_BN_steps, with_BN_train_losses)\r\n",
"plt.title('Train Loss')\r\n",
"plt.xlabel('Step')\r\n",
"plt.ylabel('Loss')\r\n",
"plt.legend(['w/o BN', 'w/ BN'])\r\n",
"plt.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
}
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 573
},
"id": "oed-_EOmnLoQ",
"outputId": "fbe5b165-89dd-40ad-8849-ec8a3d768c8c"
},
"source": [
"plt.plot([i for i in range(len(without_BN_test_accuracies))], without_BN_test_accuracies)\r\n",
"plt.plot([i for i in range(len(with_BN_test_accuracies))], with_BN_test_accuracies)\r\n",
"plt.title('Test Accuracy')\r\n",
"plt.xlabel('Epoch')\r\n",
"plt.ylabel('Accuracy')\r\n",
"plt.legend(['w/o BN', 'w/ BN'])\r\n",
"plt.show()\r\n",
"\r\n",
"plt.plot([i for i in range(len(without_BN_test_losses))], without_BN_test_losses)\r\n",
"plt.plot([i for i in range(len(with_BN_test_losses))], with_BN_test_losses)\r\n",
"plt.title('Test Loss')\r\n",
"plt.xlabel('Epoch')\r\n",
"plt.ylabel('Loss')\r\n",
"plt.legend(['w/o BN', 'w/ BN'])\r\n",
"plt.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
},
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
}
]
}
]
}