{"cells":[{"metadata":{"_uuid":"1382c63fe24710d3b2840e7dcf172cddbf533743","_cell_guid":"a0bf0fa7-c527-4fd3-b504-02a88fc94798","trusted":true},"cell_type":"code","source":"# Import Libraries\nimport torch\nimport torch.nn as nn\nfrom torch.autograd import Variable\nfrom torch.utils.data import DataLoader\nimport pandas as pd\nimport numpy as np\nfrom sklearn.model_selection import train_test_split\nimport matplotlib.pyplot as plt\nimport random","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## Load Dataset","execution_count":null},{"metadata":{"_uuid":"c6e0d7d3843719091564a580dbe08f67ee0d93ec","_cell_guid":"59cdc9d5-da8f-4d7a-abc5-c62b0008afb0","trusted":true},"cell_type":"code","source":"# Prepare Dataset\n# load data\ntrain = pd.read_csv(r\"../input/train.csv\",dtype = np.float32)\n\n# split data into features(pixels) and labels(numbers from 0 to 9)\ntargets_numpy = train.label.values\nfeatures_numpy = train.loc[:,train.columns != \"label\"].values/255 # normalization\n\n\n# train test split. Size of train data is 80% and size of test data is 20%. \nfeatures_train, features_test, targets_train, targets_test = train_test_split(features_numpy,\n                                                                             targets_numpy,\n                                                                             test_size = 0.99,\n                                                                             random_state = 42) \n\n# create feature and targets tensor for train set. As you remember we need variable to accumulate gradients. Therefore first we create tensor, then we will create variable\nfeaturesTrain = torch.from_numpy(features_train)\ntargetsTrain = torch.from_numpy(targets_train).type(torch.LongTensor) # data type is long\n\n# create feature and targets tensor for test set.\nfeaturesTest = torch.from_numpy(features_test)[:1000]\ntargetsTest = torch.from_numpy(targets_test).type(torch.LongTensor)[:1000] # data type is long\n\n# batch_size, epoch and iteration\nbatch_size = 100\nn_iters = 10000\nnum_epochs = n_iters / (len(features_train) / batch_size)\nnum_epochs = int(num_epochs)\n\n# Pytorch train and test sets\ntrain = torch.utils.data.TensorDataset(featuresTrain,targetsTrain)\ntest = torch.utils.data.TensorDataset(featuresTest,targetsTest)\n\n# data loader\ntrain_loader = DataLoader(train, batch_size = batch_size, shuffle = False)\ntest_loader = DataLoader(test, batch_size = batch_size, shuffle = False)\n\n# visualize one of the images in data set\nplt.imshow(features_numpy[10].reshape(28,28))\nplt.axis(\"off\")\nplt.title(str(targets_numpy[10]))\nplt.savefig('graph.png')\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"len(train_loader)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"len(test_loader)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## Vanila DNN","execution_count":null},{"metadata":{},"cell_type":"markdown","source":"### 1. Creating Network","execution_count":null},{"metadata":{"_uuid":"cf25ee4b28129a47bac4c9dc7d932295155b79f7","_cell_guid":"6925f8ed-54b7-4d9a-9801-acd65f213bc9","trusted":true},"cell_type":"code","source":"# Import Libraries\nimport torch\nimport torch.nn as nn\nfrom torch.autograd import Variable","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"\"\"\"\n# pytorch template\nclass cls_name(nn.Module):\n    def __init__(self, hp1, hp2, .....)\n        super(cls_name, self).__init__()\n        # define network components\n        # self.layer_ke_Satu = nn.Linear(in_size, out_size)\n        \n    def forward(self, x):\n        # define network behaviour\n        return self.layer_ke_Satu(x) \n\"\"\";","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Create ANN Model\nclass ANNModel(nn.Module):\n    def __init__(self, input_dim, hidden_dim, output_dim):\n        super(ANNModel, self).__init__()\n        self.layer_1 = nn.Linear(input_dim, hidden_dim)\n        self.ac_1 = nn.ReLU()\n        \n        self.layer_2 = nn.Linear(hidden_dim, hidden_dim)\n        self.ac_2 = nn.ReLU()\n        \n        self.layer_3 = nn.Linear(hidden_dim, hidden_dim)\n        self.ac_3 = nn.ReLU()\n        \n        self.layer_4 = nn.Linear(hidden_dim, output_dim)\n\n    def forward(self, x):\n        out = self.layer_1(x)\n        out = self.ac_1(out)\n        \n        out = self.layer_2(out)\n        out = self.ac_2(out)\n        \n        out = self.layer_3(out)\n        out = self.ac_3(out)\n        \n        out = self.layer_4(out)\n        return out","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"### 2. Hyper Parameters","execution_count":null},{"metadata":{"_uuid":"cefd0bb2f23b80f30ca65cbb08859ad81ab12e08","_cell_guid":"3472f1c1-5888-4abe-822c-3a493a5f8be5","trusted":true},"cell_type":"code","source":"# instantiate ANN\ninput_dim = 28*28\nhidden_dim = 30 #hidden layer dim is one of the hyper parameter and it should be chosen and tuned. For now I only say 150 there is no reason.\noutput_dim = 10\n\n# Create ANN\nmodel = ANNModel(input_dim, hidden_dim, output_dim)\n\n# Cross Entropy Loss \nerror = nn.CrossEntropyLoss()\n\n# SGD Optimizer\nlearning_rate = 0.02\noptimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"### 3. Training","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"# training step :\n# 1 get input\n# 2 clear gradient\n# 3 get model output \n# 4 calculate loss\n# 5 backprop loss\n# repeat","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"c91694f3af94e4e1b76ab01489e186718c70ccd3","_cell_guid":"7550e98b-5011-4d09-88ee-97b0ecbc6f19","trusted":true},"cell_type":"code","source":"# ANN model training\ncount = 0\nloss_list = []\niteration_list = []\naccuracy_list = []\nfor epoch in range(num_epochs):\n    # for every epoch pass the batched data\n    for i, (images, labels) in enumerate(train_loader):\n        # 1 get input\n        train = Variable(images.view(-1, 28*28))\n        labels = Variable(labels)\n        \n        # 2 clear gradients\n        optimizer.zero_grad()\n        \n        # 3 get model output\n        outputs = model(train)\n        \n        # 4 Calculate loss\n        loss = error(outputs, labels)\n        \n        # 5 backprop loss\n        # 5.1 Calculating gradients\n        loss.backward()\n        # 5.2 Update parameters\n        optimizer.step()\n        \n        # step var\n        count += 1\n        \n        if count % 50 == 0:\n            # Calculate Accuracy         \n            correct = 0\n            total = 0\n            # Predict test dataset\n            model.eval()\n            for images, labels in test_loader:\n\n                test = Variable(images.view(-1, 28*28))\n                \n                # Forward propagation\n                outputs = model(test)\n                \n                # Get predictions from the maximum value\n                predicted = torch.max(outputs.data, 1)[1]\n                \n                # Total number of labels\n                total += len(labels)\n\n                # Total correct predictions\n                correct += (predicted == labels).sum()\n            \n            accuracy = 100 * correct / float(total)\n            model.train()\n            # store loss and iteration\n            loss_list.append(loss.data)\n            iteration_list.append(count)\n            accuracy_list.append(accuracy)\n            \n        if count % 500 == 0:\n            # Print Loss\n            print('Iteration: {}  Loss: {}  Accuracy: {} %'.format(count, loss.data, accuracy))","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"c5e2e6da7f1ee801e38358dc28d4c99e32d2b761","_cell_guid":"5579a7d6-7766-4d0f-b9d0-584cb4f28321","trusted":true},"cell_type":"code","source":"# visualization loss \nplt.plot(iteration_list,loss_list)\nplt.xlabel(\"Number of iteration\")\nplt.ylabel(\"Loss\")\nplt.title(\"ANN: Loss vs Number of iteration\")\nplt.show()\n\n# visualization accuracy \nplt.plot(iteration_list,accuracy_list,color = \"red\")\nplt.xlabel(\"Number of iteration\")\nplt.ylabel(\"Accuracy\")\nplt.title(\"ANN: Accuracy vs Number of iteration\")\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"### DNN + L2 (Manual)","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"# instantiate ANN\ninput_dim = 28*28\nhidden_dim = 30 #hidden layer dim is one of the hyper parameter and it should be chosen and tuned. For now I only say 150 there is no reason.\noutput_dim = 10\n\n# Create ANN\nmodel = ANNModel(input_dim, hidden_dim, output_dim)\n\n# Cross Entropy Loss \nerror = nn.CrossEntropyLoss()\nreg_lambda = 0.0001\n\n# SGD Optimizer\nlearning_rate = 0.02\noptimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# ANN model training\ncount = 0\nloss_list = []\niteration_list = []\naccuracy_list = []\nfor epoch in range(num_epochs):\n    for i, (images, labels) in enumerate(train_loader):\n        # 1 get input\n        train = Variable(images.view(-1, 28*28))\n        labels = Variable(labels)\n        \n        # 2 clear gradients\n        optimizer.zero_grad()\n        \n        # 3 get model output\n        outputs = model(train)\n        \n        # 4 calcilate loss\n        # Calculate softmax and ross entropy loss + L2\n        ##############\n        reg = 0\n        for param in model.parameters():\n          reg += 0.5 * (param ** 2).sum()  # you can replace it with abs().sum() to get L1 regularization\n        #############\n        loss = error(outputs, labels) + (reg_lambda * reg) \n        \n        # 5 backprop\n        # 5.1 Calculating gradients\n        loss.backward()\n        # 5.2 Update parameters\n        optimizer.step()\n        \n        # step var\n        count += 1\n        \n        if count % 50 == 0:\n            # Calculate Accuracy         \n            correct = 0\n            total = 0\n            # Predict test dataset\n            for images, labels in test_loader:\n\n                test = Variable(images.view(-1, 28*28))\n                \n                # Forward propagation\n                outputs = model(test)\n                \n                # Get predictions from the maximum value\n                predicted = torch.max(outputs.data, 1)[1]\n                \n                # Total number of labels\n                total += len(labels)\n\n                # Total correct predictions\n                correct += (predicted == labels).sum()\n            \n            accuracy = 100 * correct / float(total)\n            \n            # store loss and iteration\n            loss_list.append(loss.data)\n            iteration_list.append(count)\n            accuracy_list.append(accuracy)\n            \n        if count % 500 == 0:\n            # Print Loss\n            print('Iteration: {}  Loss: {}  Accuracy: {} %'.format(count, loss.data, accuracy))","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"### DNN + L2 (Auto)","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"# instantiate ANN\ninput_dim = 28*28\nhidden_dim = 150 #hidden layer dim is one of the hyper parameter and it should be chosen and tuned. For now I only say 150 there is no reason.\noutput_dim = 10\n\n# Create ANN\nmodel = ANNModel(input_dim, hidden_dim, output_dim)\n\n# Cross Entropy Loss \nerror = nn.CrossEntropyLoss()\nreg_lambda = 0.0002\n\n# SGD Optimizer\nlearning_rate = 0.02\noptimizer = torch.optim.SGD(model.parameters(), lr=learning_rate, weight_decay=reg_lambda)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# ANN model training\ncount = 0\nloss_list = []\niteration_list = []\naccuracy_list = []\nfor epoch in range(num_epochs):\n    for i, (images, labels) in enumerate(train_loader):\n\n        train = Variable(images.view(-1, 28*28))\n        labels = Variable(labels)\n        \n        # Clear gradients\n        optimizer.zero_grad()\n        \n        # Forward propagation\n        outputs = model(train)\n        \n        # Calculate softmax and ross entropy loss\n        loss = error(outputs, labels)\n        \n        # Calculating gradients\n        loss.backward()\n        \n        # Update parameters\n        optimizer.step()\n        \n        count += 1\n        \n        if count % 50 == 0:\n            # Calculate Accuracy         \n            correct = 0\n            total = 0\n            # Predict test dataset\n            for images, labels in test_loader:\n\n                test = Variable(images.view(-1, 28*28))\n                \n                # Forward propagation\n                outputs = model(test)\n                \n                # Get predictions from the maximum value\n                predicted = torch.max(outputs.data, 1)[1]\n                \n                # Total number of labels\n                total += len(labels)\n\n                # Total correct predictions\n                correct += (predicted == labels).sum()\n            \n            accuracy = 100 * correct / float(total)\n            \n            # store loss and iteration\n            loss_list.append(loss.data)\n            iteration_list.append(count)\n            accuracy_list.append(accuracy)\n        if count % 500 == 0:\n            # Print Loss\n            print('Iteration: {}  Loss: {}  Accuracy: {} %'.format(count, loss.data, accuracy))","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"### DNN + DO","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"# Create ANN Model\nclass ANNModel(nn.Module):\n    def __init__(self, input_dim, hidden_dim, output_dim, dropout_rate):\n        super(ANNModel, self).__init__()\n\n        # Linear function 1: 784 --> 150\n        self.fc1 = nn.Linear(input_dim, hidden_dim) \n        self.relu1 = nn.ReLU()\n        \n        # Linear function 2: 150 --> 150\n        self.fc2 = nn.Linear(hidden_dim, hidden_dim)\n        self.tanh2 = nn.Tanh()\n        \n        # Linear function 3: 150 --> 150\n        self.fc3 = nn.Linear(hidden_dim, hidden_dim)\n        self.elu3 = nn.ELU()\n        \n        # Linear function 4 (readout): 150 --> 10\n        self.fc4 = nn.Linear(hidden_dim, output_dim)\n        \n        # dropout\n        self.dropout = nn.Dropout(dropout_rate)\n    \n    def forward(self, x):\n        # Linear function 1\n        out = self.fc1(x)\n        out = self.relu1(out)\n        out = self.dropout(out)\n        \n        # Linear function 2\n        out = self.fc2(out)\n        out = self.tanh2(out)\n        out = self.dropout(out)\n        \n        # Linear function 2\n        out = self.fc3(out)\n        out = self.elu3(out)\n        out = self.dropout(out)\n        \n        # Linear function 4 (readout)\n        out = self.fc4(out)\n        return out","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# instantiate ANN\ninput_dim = 28*28\nhidden_dim = 150 #hidden layer dim is one of the hyper parameter and it should be chosen and tuned. For now I only say 150 there is no reason.\noutput_dim = 10\ndropout_rate = 0.2\n\n# Create ANN\nmodel = ANNModel(input_dim, hidden_dim, output_dim, dropout_rate)\n\n# Cross Entropy Loss \nerror = nn.CrossEntropyLoss()\n\n# SGD Optimizer\nlearning_rate = 0.02\noptimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# ANN model training\ncount = 0\nloss_list = []\niteration_list = []\naccuracy_list = []\nfor epoch in range(num_epochs):\n    for i, (images, labels) in enumerate(train_loader):\n        train = Variable(images.view(-1, 28*28))\n        labels = Variable(labels)\n        \n        # Clear gradients\n        optimizer.zero_grad()\n        \n        # Forward propagation\n        outputs = model(train)\n        \n        # Calculate softmax and ross entropy loss\n        loss = error(outputs, labels)\n        \n        # Calculating gradients\n        loss.backward()\n        \n        # Update parameters\n        optimizer.step()\n        \n        count += 1\n        \n        if count % 50 == 0:\n            # Calculate Accuracy         \n            correct = 0\n            total = 0\n            # Predict test dataset\n            model.eval()\n            for images, labels in test_loader:\n\n                test = Variable(images.view(-1, 28*28))\n                \n                # Forward propagation\n                outputs = model(test)\n                \n                # Get predictions from the maximum value\n                predicted = torch.max(outputs.data, 1)[1]\n                \n                # Total number of labels\n                total += len(labels)\n\n                # Total correct predictions\n                correct += (predicted == labels).sum()\n            \n            accuracy = 100 * correct / float(total)\n            model.train()\n            \n            # store loss and iteration\n            loss_list.append(loss.data)\n            iteration_list.append(count)\n            accuracy_list.append(accuracy)\n        if count % 500 == 0:\n            # Print Loss\n            print('Iteration: {}  Loss: {}  Accuracy: {} %'.format(count, loss.data, accuracy))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}