code/models/lstm_pgd.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# LSTM With Attention on PGD Dataset"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Imports and Loading Data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "Laf1vf7gkUcB"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from torch.utils.data import DataLoader, TensorDataset"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "opEBeEDXlhLk",
        "outputId": "55744c3a-7602-4a59-f8f6-a4d9677bd9ae"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Using device: mps\n"
          ]
        }
      ],
      "source": [
        "# Check if GPU is available\n",
        "device = torch.device(\"mps\" if torch.backends.mps.is_available() else \"cpu\")\n",
        "print(f\"Using device: {device}\")\n",
        "\n",
        "# Load adversarial data\n",
        "adv_train = pd.read_csv(\"/Users/aryankargwal/adverscredit/models/data/pgd_train.csv\")\n",
        "adv_test = pd.read_csv(\"/Users/aryankargwal/adverscredit/models/data/pgd_test.csv\")\n",
        "\n",
        "# Assuming the last column is the label and the rest are features\n",
        "x_train_adv = adv_train.iloc[:, :-1].values  # Features\n",
        "y_train_adv = adv_train.iloc[:, -1].values   # Labels\n",
        "x_test_adv = adv_test.iloc[:, :-1].values    # Features\n",
        "y_test_adv = adv_test.iloc[:, -1].values     # Labels\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "r0Cp1gdnlxaI"
      },
      "outputs": [],
      "source": [
        "# Normalize features\n",
        "scaler = StandardScaler()\n",
        "x_train_adv = scaler.fit_transform(x_train_adv)\n",
        "x_test_adv = scaler.transform(x_test_adv)\n",
        "\n",
        "# Reshape for LSTM: (num_samples, seq_length, num_features)\n",
        "x_train_adv = x_train_adv.reshape((x_train_adv.shape[0], 1, x_train_adv.shape[1]))\n",
        "x_test_adv = x_test_adv.reshape((x_test_adv.shape[0], 1, x_test_adv.shape[1]))\n",
        "\n",
        "# Convert numpy arrays to PyTorch tensors\n",
        "x_train_adv_tensor = torch.tensor(x_train_adv, dtype=torch.float32).to(device)\n",
        "x_test_adv_tensor = torch.tensor(x_test_adv, dtype=torch.float32).to(device)\n",
        "y_train_adv_tensor = torch.tensor(y_train_adv, dtype=torch.float32).to(device)\n",
        "y_test_adv_tensor = torch.tensor(y_test_adv, dtype=torch.float32).to(device)\n",
        "\n",
        "# Create DataLoader for adversarial data\n",
        "train_adv_dataset = TensorDataset(x_train_adv_tensor, y_train_adv_tensor)\n",
        "test_adv_dataset = TensorDataset(x_test_adv_tensor, y_test_adv_tensor)\n",
        "\n",
        "train_adv_loader = DataLoader(train_adv_dataset, batch_size=64, shuffle=True)\n",
        "test_adv_loader = DataLoader(test_adv_dataset, batch_size=64, shuffle=False)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Defining LSTM with Attention"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "I20EYA-klxyd"
      },
      "outputs": [],
      "source": [
        "class Attention(nn.Module):\n",
        "    def __init__(self, hidden_dim):\n",
        "        super(Attention, self).__init__()\n",
        "        self.W = nn.Linear(hidden_dim, hidden_dim)\n",
        "        self.v = nn.Parameter(torch.randn(hidden_dim))\n",
        "\n",
        "    def forward(self, hidden_states):\n",
        "        scores = torch.tanh(self.W(hidden_states))\n",
        "        scores = torch.matmul(scores, self.v.unsqueeze(0).unsqueeze(-1)).squeeze(-1)\n",
        "        attn_weights = torch.softmax(scores, dim=1)\n",
        "        context = torch.bmm(attn_weights.unsqueeze(1), hidden_states).squeeze(1)\n",
        "        return context\n",
        "\n",
        "class LSTMWithAttention(nn.Module):\n",
        "    def __init__(self, input_dim, hidden_dim, output_dim, num_layers=1):\n",
        "        super(LSTMWithAttention, self).__init__()\n",
        "        self.lstm = nn.LSTM(input_dim, hidden_dim, num_layers, batch_first=True)\n",
        "        self.attention = Attention(hidden_dim)\n",
        "        self.fc = nn.Linear(hidden_dim, output_dim)\n",
        "\n",
        "    def forward(self, x):\n",
        "        lstm_out, _ = self.lstm(x)\n",
        "        attn_out = self.attention(lstm_out)\n",
        "        out = self.fc(attn_out)\n",
        "        return out"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Training Model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UAmxQHr5l9Dd",
        "outputId": "ac64daa0-97eb-4628-998c-1419e2fdce88"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/10, Loss: 0.0282, Accuracy: 0.9912\n",
            "Epoch 2/10, Loss: 0.0126, Accuracy: 0.9955\n",
            "Epoch 3/10, Loss: 0.0093, Accuracy: 0.9965\n",
            "Epoch 4/10, Loss: 0.0071, Accuracy: 0.9974\n",
            "Epoch 5/10, Loss: 0.0056, Accuracy: 0.9980\n",
            "Epoch 6/10, Loss: 0.0045, Accuracy: 0.9984\n",
            "Epoch 7/10, Loss: 0.0037, Accuracy: 0.9987\n",
            "Epoch 8/10, Loss: 0.0031, Accuracy: 0.9990\n",
            "Epoch 9/10, Loss: 0.0026, Accuracy: 0.9992\n",
            "Epoch 10/10, Loss: 0.0022, Accuracy: 0.9993\n",
            "Training complete.\n"
          ]
        }
      ],
      "source": [
        "# Model parameters\n",
        "input_dim = x_train_adv_tensor.shape[2]  # Number of features\n",
        "hidden_dim = 128\n",
        "output_dim = 1  # Binary classification\n",
        "num_layers = 1\n",
        "\n",
        "# Initialize the model\n",
        "model = LSTMWithAttention(input_dim, hidden_dim, output_dim, num_layers).to(device)\n",
        "\n",
        "def train_model(model, train_loader, num_epochs=10, learning_rate=0.001):\n",
        "    criterion = nn.BCEWithLogitsLoss()\n",
        "    optimizer = optim.Adam(model.parameters(), lr=learning_rate)\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        model.train()\n",
        "        running_loss = 0.0\n",
        "        correct = 0\n",
        "        total = 0\n",
        "\n",
        "        for inputs, labels in train_loader:\n",
        "            optimizer.zero_grad()\n",
        "            outputs = model(inputs)\n",
        "            loss = criterion(outputs.squeeze(), labels)\n",
        "            loss.backward()\n",
        "            optimizer.step()\n",
        "\n",
        "            running_loss += loss.item() * inputs.size(0)\n",
        "\n",
        "            # Compute predictions and accuracy\n",
        "            predicted = (torch.sigmoid(outputs.squeeze()) > 0.5).float()\n",
        "            correct += (predicted == labels).sum().item()\n",
        "            total += labels.size(0)\n",
        "\n",
        "        epoch_loss = running_loss / len(train_loader.dataset)\n",
        "        epoch_accuracy = correct / total\n",
        "        print(f\"Epoch {epoch + 1}/{num_epochs}, Loss: {epoch_loss:.4f}, Accuracy: {epoch_accuracy:.4f}\")\n",
        "\n",
        "    print(\"Training complete.\")\n",
        "\n",
        "# Training the model on adversarial data\n",
        "train_model(model, train_adv_loader)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Evaluate Model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 715
        },
        "id": "2ZRg1phxmA2M",
        "outputId": "950ac748-a341-4b40-82e4-5c06df8545cd"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Confusion Matrix:\n",
            "[[56696    54]\n",
            " [   34 56942]]\n",
            "\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "         0.0       1.00      1.00      1.00     56750\n",
            "         1.0       1.00      1.00      1.00     56976\n",
            "\n",
            "    accuracy                           1.00    113726\n",
            "   macro avg       1.00      1.00      1.00    113726\n",
            "weighted avg       1.00      1.00      1.00    113726\n",
            "\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.metrics import confusion_matrix, classification_report, roc_curve, auc\n",
        "\n",
        "def evaluate_model(model, test_loader):\n",
        "    model.eval()\n",
        "    all_labels = []\n",
        "    all_preds = []\n",
        "\n",
        "    with torch.no_grad():\n",
        "        for inputs, labels in test_loader:\n",
        "            outputs = model(inputs)\n",
        "            predicted = (torch.sigmoid(outputs.squeeze()) > 0.5).float()\n",
        "            all_labels.extend(labels.cpu().numpy())\n",
        "            all_preds.extend(predicted.cpu().numpy())\n",
        "\n",
        "    all_labels = np.array(all_labels)\n",
        "    all_preds = np.array(all_preds)\n",
        "\n",
        "    # Compute metrics\n",
        "    conf_matrix = confusion_matrix(all_labels, all_preds)\n",
        "    class_report = classification_report(all_labels, all_preds)\n",
        "\n",
        "    # ROC Curve\n",
        "    fpr, tpr, _ = roc_curve(all_labels, all_preds)\n",
        "    roc_auc = auc(fpr, tpr)\n",
        "\n",
        "    # Print confusion matrix and classification report\n",
        "    print(\"Confusion Matrix:\")\n",
        "    print(conf_matrix)\n",
        "    print(\"\\nClassification Report:\")\n",
        "    print(class_report)\n",
        "\n",
        "    # Plot ROC Curve\n",
        "    plt.figure()\n",
        "    plt.plot(fpr, tpr, color='darkorange', lw=2, label='ROC curve (area = %0.2f)' % roc_auc)\n",
        "    plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')\n",
        "    plt.xlim([0.0, 1.0])\n",
        "    plt.ylim([0.0, 1.05])\n",
        "    plt.xlabel('False Positive Rate')\n",
        "    plt.ylabel('True Positive Rate')\n",
        "    plt.title('Receiver Operating Characteristic (ROC)')\n",
        "    plt.legend(loc='lower right')\n",
        "    plt.show()\n",
        "\n",
        "# Use the evaluate_model function with adversarial test data\n",
        "model_path = \"/Users/aryankargwal/adverscredit/models/weights/lst_pgd.pth\"\n",
        "model.load_state_dict(torch.load(model_path, map_location=torch.device('mps'))) \n",
        "evaluate_model(model, test_adv_loader)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hBvthMYbmd7h",
        "outputId": "925c8bd6-208a-4e1c-bde9-232874e65fc2"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model saved to lst_pgd.pth\n"
          ]
        }
      ],
      "source": [
        "# Save the model\n",
        "model_path = \"lst_pgd.pth\"\n",
        "torch.save(model.state_dict(), model_path)\n",
        "print(f\"Model saved to {model_path}\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "kLbenoIQn1Ck"
      },
      "outputs": [],
      "source": []
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "gpuType": "T4",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3.11.0 ('base')",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.11.0"
    },
    "vscode": {
      "interpreter": {
        "hash": "40d3a090f54c6569ab1632332b64b2c03c39dcf918b08424e98f38b5ae0af88f"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}