fashion-mnist-classifier.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 52,
      "id": "616b1bd3",
      "metadata": {
        "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
        "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
        "execution": {
          "iopub.execute_input": "2023-05-04T02:15:49.891829Z",
          "iopub.status.busy": "2023-05-04T02:15:49.891477Z",
          "iopub.status.idle": "2023-05-04T02:15:53.784960Z",
          "shell.execute_reply": "2023-05-04T02:15:53.784059Z",
          "shell.execute_reply.started": "2023-05-04T02:15:49.891797Z"
        },
        "papermill": {
          "duration": 3.352339,
          "end_time": "2023-01-20T22:38:35.620981",
          "exception": false,
          "start_time": "2023-01-20T22:38:32.268642",
          "status": "completed"
        },
        "tags": [],
        "id": "616b1bd3"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import torch\n",
        "import matplotlib.pyplot as plt\n",
        "from torch import nn\n",
        "from torch.utils.data import DataLoader\n",
        "from torch.utils.data import Subset\n",
        "from torchvision import datasets\n",
        "from torchvision.transforms import ToTensor\n",
        "\n",
        "plt.rc('image', cmap='Greys')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 53,
      "id": "db441508-e51f-4168-9eac-564b40591591",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:15:53.787560Z",
          "iopub.status.busy": "2023-05-04T02:15:53.786934Z",
          "iopub.status.idle": "2023-05-04T02:15:53.819534Z",
          "shell.execute_reply": "2023-05-04T02:15:53.818563Z",
          "shell.execute_reply.started": "2023-05-04T02:15:53.787533Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "db441508-e51f-4168-9eac-564b40591591",
        "outputId": "0d8674f6-bd7f-4a4f-dc28-ed67846869ba"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Using cuda device\n"
          ]
        }
      ],
      "source": [
        "device = \"cuda\" if torch.cuda.is_available() else \"mps\" if torch.backends.mps.is_available() else \"cpu\"\n",
        "print(f\"Using {device} device\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 54,
      "id": "0c1094f0-f591-49b9-bc74-ca9e78be9e6a",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:15:53.820919Z",
          "iopub.status.busy": "2023-05-04T02:15:53.820615Z",
          "iopub.status.idle": "2023-05-04T02:15:53.826866Z",
          "shell.execute_reply": "2023-05-04T02:15:53.825814Z",
          "shell.execute_reply.started": "2023-05-04T02:15:53.820891Z"
        },
        "id": "0c1094f0-f591-49b9-bc74-ca9e78be9e6a"
      },
      "outputs": [],
      "source": [
        "classes = [\n",
        "    \"T-shirt/top\",\n",
        "    \"Trouser\",\n",
        "    \"Pullover\",\n",
        "    \"Dress\",\n",
        "    \"Coat\",\n",
        "    \"Sandal\",\n",
        "    \"Shirt\",\n",
        "    \"Sneaker\",\n",
        "    \"Bag\",\n",
        "    \"Ankle boot\",\n",
        "]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 55,
      "id": "e5f6c8bd-0599-4934-9c3d-649c900c84d9",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:15:53.829287Z",
          "iopub.status.busy": "2023-05-04T02:15:53.829040Z",
          "iopub.status.idle": "2023-05-04T02:15:54.145829Z",
          "shell.execute_reply": "2023-05-04T02:15:54.145020Z",
          "shell.execute_reply.started": "2023-05-04T02:15:53.829263Z"
        },
        "id": "e5f6c8bd-0599-4934-9c3d-649c900c84d9"
      },
      "outputs": [],
      "source": [
        "train_data = datasets.FashionMNIST(\n",
        "    root=\"data\",\n",
        "    train=True,\n",
        "    download=True,\n",
        "    transform=ToTensor()\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 56,
      "id": "53b16342-a329-414c-b711-6d4717a92ead",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:15:54.152051Z",
          "iopub.status.busy": "2023-05-04T02:15:54.149925Z",
          "iopub.status.idle": "2023-05-04T02:16:08.128691Z",
          "shell.execute_reply": "2023-05-04T02:16:08.127861Z",
          "shell.execute_reply.started": "2023-05-04T02:15:54.152024Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "53b16342-a329-414c-b711-6d4717a92ead",
        "outputId": "31694380-8b20-490e-fb2c-065f666ce5fc"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "All good!\n"
          ]
        }
      ],
      "source": [
        "train_data_by_label = [[] for i in range(10)]\n",
        "\n",
        "for x, y in train_data:\n",
        "    train_data_by_label[y].append(x)\n",
        "\n",
        "for i in range(len(train_data_by_label)):\n",
        "    if len(train_data_by_label[i]) != 6000:\n",
        "        print('Error. Unexpected length.')\n",
        "print('All good!')"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "8bc997f2-0886-40e7-917c-f6477eba7bf8",
      "metadata": {
        "id": "8bc997f2-0886-40e7-917c-f6477eba7bf8"
      },
      "source": [
        "## Generating mean items\n",
        "\n",
        "The mean items are created by getting the mean of all the tensors in each category of the training dataset.\n",
        "\n",
        "It takes the average of each pixel for all the images in each category."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 57,
      "id": "46d4a9b2-e3e4-4bbe-b396-5bb12007c11c",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:08.130691Z",
          "iopub.status.busy": "2023-05-04T02:16:08.130194Z"
        },
        "id": "46d4a9b2-e3e4-4bbe-b396-5bb12007c11c"
      },
      "outputs": [],
      "source": [
        "mean_items = [None for i in range(10)]\n",
        "samples = [None for i in range(10)]\n",
        "sample_index = 0\n",
        "\n",
        "for label in range(10):\n",
        "    stacked = torch.stack(train_data_by_label[label]).squeeze()\n",
        "    samples[label] = stacked[sample_index]\n",
        "    mean = stacked.mean(0)\n",
        "    mean_items[label] = mean"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "3cd8c702-2cea-4786-adaa-559111497a53",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-03-27T22:13:54.829821Z",
          "iopub.status.busy": "2023-03-27T22:13:54.828911Z",
          "iopub.status.idle": "2023-03-27T22:13:54.832994Z",
          "shell.execute_reply": "2023-03-27T22:13:54.832245Z",
          "shell.execute_reply.started": "2023-03-27T22:13:54.829788Z"
        },
        "id": "3cd8c702-2cea-4786-adaa-559111497a53"
      },
      "source": [
        "## Try it yourself:\n",
        "Change `sample_index` to see different samples"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "7bb8c1ea-bfc0-46b8-84d2-17e85a6c5417",
      "metadata": {
        "id": "7bb8c1ea-bfc0-46b8-84d2-17e85a6c5417"
      },
      "source": [
        "# Samples"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 58,
      "id": "8e05fdc7-7860-4a6e-93a1-95d9c91b5221",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:08.643226Z",
          "iopub.status.busy": "2023-05-04T02:16:08.642964Z",
          "iopub.status.idle": "2023-05-04T02:16:09.448273Z",
          "shell.execute_reply": "2023-05-04T02:16:09.447492Z",
          "shell.execute_reply.started": "2023-05-04T02:16:08.643200Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 156
        },
        "id": "8e05fdc7-7860-4a6e-93a1-95d9c91b5221",
        "outputId": "13353cfa-7103-4dd9-be56-6a1bd51a3494"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 2000x1000 with 10 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "fig, axs = plt.subplots(1, 10, figsize=(20,10))\n",
        "axs = axs.ravel()\n",
        "\n",
        "for i in range(10):\n",
        "    axs[i].imshow(samples[i])\n",
        "    axs[i].set_title(f'{classes[i]}')\n",
        "    axs[i].axis('off')\n",
        "    \n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "ecb61858-2db2-418a-9d43-d1508bd1dfda",
      "metadata": {
        "id": "ecb61858-2db2-418a-9d43-d1508bd1dfda"
      },
      "source": [
        "# Mean items"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 59,
      "id": "28950126-7e3d-4529-bc33-3d112ab36385",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:09.456837Z",
          "iopub.status.busy": "2023-05-04T02:16:09.449616Z",
          "iopub.status.idle": "2023-05-04T02:16:10.432067Z",
          "shell.execute_reply": "2023-05-04T02:16:10.427416Z",
          "shell.execute_reply.started": "2023-05-04T02:16:09.456805Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 156
        },
        "id": "28950126-7e3d-4529-bc33-3d112ab36385",
        "outputId": "58759fa2-5877-481f-cacb-d011b71647fa"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 2000x1000 with 10 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABiEAAACtCAYAAADWI9yPAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABhF0lEQVR4nO3deXhV1b0+8DcMSSADYUiYCRBUBFQsIlQURFGqIHVAEa2CinIVq97aerX+vMq1rdehVooVpdeiFbiIFsWBQbyg1hkHrFpFQEDGMEPCFIH9+4Mnp+7vfkO+nJxDpvfzPD6Pe7HOyT57r73W2mcn600JgiCAiIiIiIiIiIiIiIhIgtWp7B0QEREREREREREREZGaSQ8hREREREREREREREQkKfQQQkREREREREREREREkkIPIUREREREREREREREJCn0EEJERERERERERERERJJCDyFERERERERERERERCQp9BBCRERERERERERERESSQg8hREREREREREREREQkKfQQQkREREREREREREREkqJWPIRYsWIFUlJS8NBDD5Vb95577kFKSsoR2CsRkSMnJSUF99xzT2z7qaeeQkpKClasWFFp+yQiIiJcRcbpkSNHon379gnfJ6m6UlJScOONN5ZbT/M/qSoO5zsaEZFEGTlyJDIzM8utd/rpp+P0009P2M89/fTT0a1bt4S9X3VVJR5CpKSkuP574403KntXQ3bt2oV77rnnkPu1detW1KtXD9OnTwcA/O53v8OLL754ZHZQEqq6tlOpnkpvEkv/S09Px9FHH40bb7wRhYWFlb17Usux9tmqVSsMHDgQf/zjH1FUVFTZuyi1yLJlyzB69Gh07NgR6enpyM7ORp8+fTBu3Djs3r07KT9z6tSpeOSRR5Ly3lJ5Pv/8cwwdOhT5+flIT09H69atcdZZZ2H8+PGVvWtSi1Vmu9S9a/WmPk2qK3uvkZKSgry8PPTv3x+zZ8+u7N2TI+yxxx5DSkoKevXqVdm7Ui1VpbG8XmXvAAA888wzoe2//vWvmDdvXqT82GOPTfq+/L//9/9w++23u+ru2rULY8eOBYAyn5DNnTsXKSkpOPvsswEcPPlDhw7F+eefn4jdlSOoKrVTqT3+67/+Cx06dMCePXvw9ttvY8KECZg1axa++OILNGzYsLJ3T2q50vb5/fffY/369XjjjTdwyy234OGHH8ZLL72E448/vrJ3UWq4V199FRdffDHS0tJw5ZVXolu3bigpKcHbb7+NX/3qV/jyyy8xceLEhP/cqVOn4osvvsAtt9yS8PeWyvHuu++if//+aNeuHa699lq0aNECq1atwvvvv49x48bh5z//eWXvotRCiW6XV1xxBS699FKkpaW56uvetfpSnyY1Qem9RhAEKCwsxFNPPYVzzz0XL7/8MgYPHlzZuydHyJQpU9C+fXt8+OGHWLp0KTp16lTZu1StVKWxvEo8hPjZz34W2n7//fcxb968SPmRUK9ePdSrd+jDcuDAAZSUlLjeb9asWejTpw9ycnISsHdSmeJtp7t27aqWXxbv3LkTGRkZlb0btd4555yDk046CQAwatQoNG3aFA8//DBmzpyJ4cOHV/LeJY/aX/Xww/YJAHfccQfmz5+PwYMHY8iQIfjqq6/QoEED+lqdY6mo5cuX49JLL0V+fj7mz5+Pli1bxv5tzJgxWLp0KV599dVK3EOpTn7729+iUaNGWLhwYWTevmHDhsrZKan1Et0u69ati7p16x6yThAE2LNnT5njt1QP6tOq7324/Iu917jmmmvQvHlz/O///q8eQtQSy5cvx7vvvosZM2Zg9OjRmDJlCu6+++7K3i2JU5VYjqmiPvroIwwcOBDNmjVDgwYN0KFDB1x99dW07sSJE1FQUIC0tDT07NkTCxcuDP07y4QoXWNzypQp6Nq1K9LS0vD4448jNzcXADB27NjYn4j9cM31AwcOYM6cORg0aFDsfXbu3Imnn346Vn/kyJGx+p9++inOOeccZGdnIzMzE2eeeSbef//90L6U/lnaW2+9hdGjR6Np06bIzs7GlVdeia1bt8Z7CCVBStd5+/jjj9G3b180bNgQv/71rwEcnOyVDprp6ek44YQT8PTTT4de/8Ybb9AlnUrXzHzqqadiZevXr8dVV12FNm3aIC0tDS1btsRPf/rTyBqvs2fPxmmnnYaMjAxkZWVh0KBB+PLLL0N1StfFW7ZsGc4991xkZWXh8ssvT9hxkcQ544wzABwcjMtap7Aia0E/9thjsX6uVatWGDNmDLZt2xb79xtvvBGZmZnYtWtX5LXDhw9HixYtsH///liZ2l/tc8YZZ+Cuu+7CypUrMXnyZACHPscHDhzAI488gq5duyI9PR3NmzfH6NGjI2OaZ6yfNm0aevTogaysLGRnZ+O4447DuHHjjswHlyPugQceQHFxMZ588snQA4hSnTp1ws033wwA2LdvH+69997YHLB9+/b49a9/jb1794ZeM3PmTAwaNAitWrVCWloaCgoKcO+994b6tdNPPx2vvvoqVq5cGZvPaf396m/ZsmXo2rUr/cWhvLy82P9PmjQJZ5xxBvLy8pCWloYuXbpgwoQJkde0b98egwcPxttvv42TTz4Z6enp6NixI/76179G6n755Zc444wz0KBBA7Rp0wa/+c1vcODAgUg9T/uUmsXbLku9+OKL6NatG9LS0tC1a1fMmTMn9O8sE6K0rc6dOxcnnXQSGjRogCeeeKLce1ep2rxtp/S7jvLaDgCsWbMGV199NZo3bx6r95e//CVUp6SkBP/5n/+JHj16oFGjRsjIyMBpp52GBQsWlLvPQRDguuuuQ2pqKmbMmBErnzx5Mnr06IEGDRqgSZMmuPTSS7Fq1arQaw91Hy41R05ODho0aBD6xeGHHnoIp5xyCpo2bYoGDRqgR48eeP755yOv3b17N2666SY0a9YMWVlZGDJkCNasWRP5Hk+qlilTpqBx48YYNGgQhg4diilTpkTq/DBjprzvfJlFixYhNzcXp59+OoqLi8ust3fvXtx9993o1KkT0tLS0LZtW9x2222R+4lD+fjjj3HKKafE7mcff/zxSB3Pd4fAwV/qu/XWW9G2bVukpaXhmGOOwUMPPYQgCGJ1qtpYXiX+EqIiNmzYgLPPPhu5ubm4/fbbkZOTgxUrVoQGrVJTp05FUVERRo8ejZSUFDzwwAO48MIL8e2336J+/fqH/Dnz58/H9OnTceONN6JZs2Y44YQTMGHCBFx//fW44IILcOGFFwJAaOmJhQsXYuPGjTj33HMBHFzOZ9SoUTj55JNx3XXXAQAKCgoAHLz5OO2005CdnY3bbrsN9evXxxNPPIHTTz8db775ZmTtsxtvvBE5OTm45557sHjxYkyYMAErV66MfYktlWfz5s0455xzcOmll+JnP/sZmjdvjt27d+P000/H0qVLceONN6JDhw547rnnMHLkSGzbti32JcnhuOiii/Dll1/i5z//Odq3b48NGzZg3rx5+O6772JfhjzzzDMYMWIEBg4ciPvvvx+7du3ChAkTcOqpp+LTTz8NfWmyb98+DBw4EKeeeioeeugh/dZIFbVs2TIAQNOmTRP+3vfccw/Gjh2LAQMG4Prrr4/1LQsXLsQ777yD+vXrY9iwYfjTn/4UWwKl1K5du/Dyyy9j5MiRsd+wU/urva644gr8+te/xmuvvYZrr70WQNnnePTo0Xjqqadw1VVX4aabbsLy5cvx6KOP4tNPP421O89YP2/ePAwfPhxnnnkm7r//fgDAV199hXfeeSeuPlaqvpdffhkdO3bEKaecUm7dUaNG4emnn8bQoUNx66234oMPPsB9992Hr776Ci+88EKs3lNPPYXMzEz84he/QGZmJubPn4///M//xI4dO/Dggw8CAO68805s374dq1evxh/+8AcAcAXcSdWWn5+P9957D1988cUhgwMnTJiArl27YsiQIahXrx5efvll3HDDDThw4ADGjBkTqrt06VIMHToU11xzDUaMGIG//OUvGDlyJHr06IGuXbsCOPhLJf3798e+fftw++23IyMjAxMnTqS/he5pn1KzeNslALz99tuYMWMGbrjhBmRlZeGPf/wjLrroInz33XflzhsXL16M4cOHY/To0bj22mtxzDHHHPLeVaq+RLedwsJC9O7dO/bQIjc3F7Nnz8Y111yDHTt2xJYn3LFjB/7nf/4Hw4cPx7XXXouioiI8+eSTGDhwID788EN0796d7sP+/ftx9dVX49lnn8ULL7wQ+0XO3/72t7jrrrtwySWXYNSoUdi4cSPGjx+Pvn374tNPPw09ZGH34VK9bd++HZs2bUIQBNiwYQPGjx+P4uLi0GoU48aNw5AhQ3D55ZejpKQE06ZNw8UXX4xXXnkl1o6Ag78UNX36dFxxxRXo3bs33nzzzdC/S9U0ZcoUXHjhhUhNTcXw4cNj30/07NkzUjee73wXLlyIgQMH4qSTTsLMmTPL/CvAAwcOYMiQIXj77bdx3XXX4dhjj8Xnn3+OP/zhD/jmm29cmQtbt27Fueeei0suuQTDhw/H9OnTcf311yM1NTX2y3Xe7w6DIMCQIUOwYMECXHPNNejevTvmzp2LX/3qV1izZk3sHqXKjeVBFTRmzJjAu2svvPBCACBYuHBhmXWWL18eAAiaNm0abNmyJVY+c+bMAEDw8ssvx8ruvvvuyM8GENSpUyf48ssvQ+UbN24MAAR33303/bl33XVXkJ+fHyrLyMgIRowYEal7/vnnB6mpqcGyZctiZWvXrg2ysrKCvn37xsomTZoUAAh69OgRlJSUxMofeOCBAEAwc+bMMo+DJBZrp/369QsABI8//nio/JFHHgkABJMnT46VlZSUBD/+8Y+DzMzMYMeOHUEQBMGCBQsCAMGCBQtCry9tw5MmTQqCIAi2bt0aAAgefPDBMvevqKgoyMnJCa699tpQ+fr164NGjRqFykeMGBEACG6//Xb355fkKr3WX3/99WDjxo3BqlWrgmnTpgVNmzYNGjRoEKxevTro169f0K9fv8hrR4wYEel7bF9V+v7Lly8PgiAINmzYEKSmpgZnn312sH///li9Rx99NAAQ/OUvfwmCIAgOHDgQtG7dOrjoootC7z99+vQAQPDWW28FQaD2V9OVtp9Djb2NGjUKTjzxxCAIyj7Hf//73wMAwZQpU0Llc+bMCZV7xvqbb745yM7ODvbt2xfvx5JqZPv27QGA4Kc//Wm5dRctWhQACEaNGhUq/+UvfxkACObPnx8r27VrV+T1o0ePDho2bBjs2bMnVjZo0KBIPyvV22uvvRbUrVs3qFu3bvDjH/84uO2224K5c+eG5ttBwNvIwIEDg44dO4bK8vPzQ+NiEBwca9PS0oJbb701VnbLLbcEAIIPPvggVK9Ro0ahcbqsn83aJ5sHSPXkbZcAgtTU1GDp0qWxss8++ywAEIwfPz5WZud/QfCvtjpnzpzIzy/r3lWqvkS3nWuuuSZo2bJlsGnTptDrL7300qBRo0ax/mnfvn3B3r17Q3W2bt0aNG/ePLj66qtjZaX3tw8++GDw/fffB8OGDQsaNGgQzJ07N1ZnxYoVQd26dYPf/va3off7/PPPg3r16oXKy7oPl+qptK+y/6WlpQVPPfVUqK4dG0tKSoJu3boFZ5xxRqzs448/DgAEt9xyS6juyJEjD/mdnlSujz76KAAQzJs3LwiCg99FtGnTJrj55ptD9Q7nO98RI0YEGRkZQRAEwdtvvx1kZ2cHgwYNCs2jgiCIfNfyzDPPBHXq1An+/ve/h+o9/vjjAYDgnXfeOeRnKe2jfv/738fK9u7dG3Tv3j3Iy8uL9c3e7w5ffPHFAEDwm9/8JvRzhg4dGqSkpIT69Ko0llf75ZhKn3y/8sor+P777w9Zd9iwYWjcuHFs+7TTTgMAfPvtt+X+nH79+qFLly6HtW+zZs1yPVndv38/XnvtNZx//vno2LFjrLxly5a47LLL8Pbbb2PHjh2h11x33XWhJ3nXX3896tWrh1mzZh3WPkripaWl4aqrrgqVzZo1Cy1atAit4V+/fn3cdNNNKC4uxptvvnlYP6NBgwZITU3FG2+8UeYyXPPmzcO2bdswfPhwbNq0KfZf3bp10atXL/onsddff/1h7Yck34ABA5Cbm4u2bdvi0ksvRWZmJl544QW0bt06oT/n9ddfR0lJCW655RbUqfOvoeHaa69FdnZ2bF31lJQUXHzxxZg1a1boTxWfffZZtG7dGqeeeioAtT85+JvhRUVFoTJ7jp977jk0atQIZ511Vqid9OjRA5mZmbF24hnrc3JysHPnTsybNy/xH0aqnNJ5UVZWVrl1S+dGv/jFL0Llt956KwCEciN++NtPRUVF2LRpE0477TTs2rULX3/9dYX3W6qus846C++99x6GDBmCzz77DA888AAGDhyI1q1b46WXXorV+2EbKf0NzX79+uHbb7/F9u3bQ+/ZpUuX2P0GAOTm5uKYY44J3XvMmjULvXv3xsknnxyqx5YlVPusfbztEjg4Z/zhbzcef/zxyM7Odt3rdujQAQMHDkz4/kvlSWTbCYIAf/vb33DeeechCILQnG3gwIHYvn07PvnkEwAHc0dSU1MBHPzN4S1btmDfvn046aSTYnV+qKSkJPZb67NmzcLZZ58d+7cZM2bgwIEDuOSSS0I/s0WLFjjqqKMi9xPsPlyqtz/96U+YN28e5s2bh8mTJ6N///4YNWpU6K+hfzg2bt26Fdu3b8dpp50Wam+ly4vdcMMNofdXQHvVNmXKFDRv3hz9+/cHcPC7iGHDhmHatGl0KcrD+c53wYIFGDhwIM4880zMmDEDaWlph9yX5557Dsceeyw6d+4c6o9Kl8v2LDlXr149jB49OradmpqK0aNHY8OGDfj4448B+L87nDVrFurWrYubbrop9DNuvfVWBEGA2bNnl7s/laHaPIQoLi7G+vXrY/9t3LgRwMGHAxdddBHGjh2LZs2a4ac//SkmTZpE1+Rq165daLu0cXqyFDp06HBY+7t+/Xp88sknrocQGzduxK5du3DMMcdE/u3YY4/FgQMHImseHnXUUaHtzMxMtGzZMpIHIEde69atYxOvUitXrsRRRx0V+nIXOHh+S//9cKSlpeH+++/H7Nmz0bx5c/Tt2xcPPPAA1q9fH6uzZMkSAAfXZ8/NzQ3999prr0UCyerVq4c2bdoc1n5I8pVOvBYsWIB//vOf+Pbbb5Nyk1jaBm0/lJqaio4dO4ba6LBhw7B79+7YDUxxcTFmzZqFiy++OLYcnNqfFBcXh74gZud4yZIl2L59O/Ly8iLtpLi4ONZOPGP9DTfcgKOPPhrnnHMO2rRpg6uvvpquZyw1Q3Z2NgBEHnQxK1euRJ06ddCpU6dQeYsWLZCTkxPq37788ktccMEFaNSoEbKzs5Gbmxv7k3/7BbPUPD179sSMGTOwdetWfPjhh7jjjjtQVFSEoUOH4p///CcA4J133sGAAQOQkZGBnJwc5ObmxtYdt23E3nsAB+8/fnjvUTpHtNh9gdpn7eRpl4CvvZXlcO91pXpIVNvZuHEjtm3bhokTJ0bma6Vf+v9wbv/000/j+OOPR3p6Opo2bYrc3Fy8+uqrtJ+677778OKLL+L555+PZN0tWbIEQRDgqKOOivzcr776KnI/we7DpXo7+eSTMWDAAAwYMACXX345Xn31VXTp0gU33ngjSkpKABz8JaXevXsjPT0dTZo0QW5uLiZMmBBqb6VzQdvX2bmhVB379+/HtGnT0L9/fyxfvhxLly7F0qVL0atXLxQWFuL//u//Iq/xfue7Z88eDBo0CCeeeCKmT5/u6jeWLFmCL7/8MtIXHX300QAQ6Y+YVq1aISMjI1RW+vrS73K93x2uXLkSrVq1ivxCVrzfMR4p1SYT4qGHHsLYsWNj2/n5+bHwkeeffx7vv/8+Xn75ZcydOxdXX301fv/73+P9998PrdFbuk65FfwgtKMsZa0LVpbZs2cjPT099sROao/DbSs/VFaeB3vKe8stt+C8887Diy++iLlz5+Kuu+7Cfffdh/nz5+PEE0+MBRo+88wzaNGiReT1PwxzAg4+2LAdnVS+k08+GSeddBL9t5SUFNp/JTugsnfv3mjfvj2mT5+Oyy67DC+//DJ2796NYcOGxeqo/dVuq1evxvbt20MTe3aODxw4gLy8PBowBhz8bWAArrE+Ly8PixYtwty5czF79mzMnj0bkyZNwpVXXkmDvKR6y87ORqtWrfDFF1+4X1NeZta2bdvQr18/ZGdn47/+679QUFCA9PR0fPLJJ/iP//gPGhQsNVNqaip69uyJnj174uijj8ZVV12F5557Dj/72c9w5plnonPnznj44YfRtm1bpKamYtasWfjDH/4QaSMVufew1D6lrHZ59913Aziy97pSvVS07ZT2Lz/72c8wYsQIWrc0G3Py5MkYOXIkzj//fPzqV79CXl4e6tati/vuuy+WbfdDAwcOxJw5c/DAAw/g9NNPR3p6euzfDhw4gJSUFMyePZvuo81jUjuu+erUqYP+/ftj3LhxWLJkCbZs2YIhQ4agb9++eOyxx9CyZUvUr18fkyZNwtSpUyt7d6UC5s+fj3Xr1mHatGmYNm1a5N+nTJkS+sspwD8OpqWl4dxzz8XMmTMxZ84cDB48uNz9OXDgAI477jg8/PDD9N/btm1b7ntINXoIceWVV8aW+QCiA0zv3r3Ru3dv/Pa3v8XUqVNx+eWXY9q0aRg1alTS9ulQN7Ovvvoq+vfvH9lP9prc3Fw0bNgQixcvjvzb119/jTp16kQa9JIlS0IPOIqLi7Fu3bpYCLZULfn5+fjHP/6BAwcOhL6EK/3T+fz8fAD/elK7bdu20OvLeopZUFCAW2+9FbfeeiuWLFmC7t274/e//z0mT54c+5PavLw8DBgwINEfSaqAxo0b0z8tjOepd2kbXLx4cWhZuJKSEixfvjzShi655BKMGzcOO3bswLPPPov27dujd+/esX9X+6vdnnnmGQAo9692CgoK8Prrr6NPnz6uG8fyxvrU1FScd955OO+883DgwAHccMMNeOKJJ3DXXXfpN51qoMGDB2PixIl477338OMf/7jMevn5+Thw4ACWLFkS++0g4GDI5rZt22L93xtvvIHNmzdjxowZ6Nu3b6ze8uXLI+9Z3gMNqTlKfxFg3bp1ePnll7F371689NJLod+28/wJflny8/Njfz34Q/a+4HDap9R8P2yXyaS+ruaJp+3k5uYiKysL+/fvL3de//zzz6Njx46YMWNGqP2UPvCwevfujX/7t3/D4MGDcfHFF+OFF16I/bJSQUEBgiBAhw4dYr8tLLJv3z4AB78D+9vf/ob09HTMnTs3tJzOpEmTQq8pnQsuX7489NeHS5cuPTI7LYdtypQpyMvLw5/+9KfIv82YMQMvvPACHn/88bgePqakpGDKlCn46U9/iosvvhizZ8+O/CWWVVBQgM8++wxnnnlm3GPj2rVrsXPnztBfQ3zzzTcAgPbt2wPwf3eYn5+P119/HUVFRaG/hrD1Sj9vVVFtfu20Y8eOsT/DGjBgAPr06QPg4J/V2Kda3bt3BwC6JFMiNWzYEED0C+Pvv/8e8+bNo0sxZWRkROrXrVsXZ599NmbOnBlaTqmwsBBTp07FqaeeGlt2oNTEiRND62JPmDAB+/btwznnnFOxDyVJce6552L9+vV49tlnY2X79u3D+PHjkZmZiX79+gE42FHUrVsXb731Vuj1jz32WGh7165d2LNnT6isoKAAWVlZsXY/cOBAZGdn43e/+x1dQ710STOpvgoKCvD111+HzuVnn32Gd95557Dfa8CAAUhNTcUf//jHUJ/65JNPYvv27ZH+bNiwYdi7dy+efvppzJkzB5dcckno39X+aq/58+fj3nvvRYcOHeia5j90ySWXYP/+/bj33nsj/7Zv377YeOkZ6zdv3hz69zp16sR+Ky/Z8wGpHLfddhsyMjIwatQoFBYWRv592bJlGDduXOwXNB555JHQv5f+JlNp/1b621M/bGslJSWRMRg4OJ/T8jc1y4IFC+hvjJdmihxzzDG0jWzfvj3yZcfhOPfcc/H+++/jww8/jJVt3Lgx8hdih9M+pebwtMtkYveuUj0ksu3UrVsXF110Ef72t7/Rv0D84bye9VUffPAB3nvvvTLff8CAAZg2bRrmzJmDK664IvaXFxdeeCHq1q2LsWPHRj5LEASRuZ/UfN9//z1ee+01pKam4thjj0XdunWRkpISWglgxYoVePHFF0OvK/3FKDtmjh8/Pun7LIdv9+7dmDFjBgYPHoyhQ4dG/rvxxhtRVFQUybc5HKmpqZgxYwZ69uyJ8847LzQPYy655BKsWbMGf/7zn+n+7ty5s9yfuW/fPjzxxBOx7ZKSEjzxxBPIzc1Fjx49APi/Ozz33HOxf/9+PProo6Gf8Yc//AEpKSmh74ar0lhebf4SoixPP/00HnvsMVxwwQUoKChAUVER/vznPyM7OzvpfxXQoEEDdOnSBc8++yyOPvpoNGnSBN26dcPGjRuxY8cO+hCiR48eeP311/Hwww+jVatW6NChA3r16oXf/OY3mDdvHk499VTccMMNqFevHp544gns3bsXDzzwQOR9SkpKcOaZZ+KSSy7B4sWL8dhjj+HUU0/FkCFDkvqZJT7XXXcdnnjiCYwcORIff/wx2rdvj+effx7vvPMOHnnkkdiTy0aNGuHiiy/G+PHjkZKSgoKCArzyyiuR9eW++eab2Pnv0qUL6tWrhxdeeAGFhYW49NJLARxcqmLChAm44oor8KMf/QiXXnopcnNz8d133+HVV19Fnz59Ih2WVC9XX301Hn74YQwcOBDXXHMNNmzYgMcffxxdu3aNhNmXJzc3F3fccQfGjh2Ln/zkJxgyZEisb+nZs2dszelSP/rRj9CpUyfceeed2Lt3b2gpJkDtr7aYPXs2vv76a+zbtw+FhYWYP38+5s2bh/z8fLz00kuhP6ln+vXrh9GjR+O+++7DokWLcPbZZ6N+/fpYsmQJnnvuOYwbNw5Dhw51jfWjRo3Cli1bcMYZZ6BNmzZYuXIlxo8fj+7du4d++11qjoKCAkydOhXDhg3DscceiyuvvBLdunVDSUkJ3n33XTz33HMYOXIkbr75ZowYMQITJ06MLWnz4Ycf4umnn8b5558f+8vSU045BY0bN8aIESNw0003ISUlBc888wz9EqdHjx549tln8Ytf/AI9e/ZEZmYmzjvvvCN9CCSBfv7zn2PXrl244IIL0Llz51g7Kv1rv6uuugqFhYWxv7gaPXo0iouL8ec//xl5eXlx/0b6bbfdhmeeeQY/+clPcPPNNyMjIwMTJ06M/SZcqcNpn1JzeNplMpV17ypVX6Lbzn//939jwYIF6NWrF6699lp06dIFW7ZswSeffILXX38dW7ZsAXDwrxRnzJiBCy64AIMGDcLy5cvx+OOPo0uXLiguLi7z/c8///zYMprZ2dl44oknUFBQgN/85je44447sGLFCpx//vnIysrC8uXL8cILL+C6667DL3/5ywodJ6naSu81gINr7k+dOhVLlizB7bffjuzsbAwaNAgPP/wwfvKTn+Cyyy7Dhg0b8Kc//QmdOnUKjaE9evTARRddhEceeQSbN29G79698eabb8Z+C70q/aa4AC+99BKKiorK/H6zd+/eyM3NxZQpUyLfQxyOBg0a4JVXXsEZZ5yBc845B2+++Sa6detG615xxRWYPn06/u3f/g0LFixAnz59sH//fnz99deYPn065s6dW+Yy2qVatWqF+++/HytWrMDRRx+NZ599FosWLcLEiRNRv359AP7vDs877zz0798fd955J1asWIETTjgBr732GmbOnIlbbrkltjIFUMXG8qAKGjNmTODdtU8++SQYPnx40K5duyAtLS3Iy8sLBg8eHHz00UexOsuXLw8ABA8++GDk9QCCu+++O7Z99913R342gGDMmDH057/77rtBjx49gtTU1Nh7/fKXvwy6dOlC63/99ddB3759gwYNGgQAghEjRoQ+y8CBA4PMzMygYcOGQf/+/YN333039PpJkyYFAII333wzuO6664LGjRsHmZmZweWXXx5s3ry5vMMlCcTaab9+/YKuXbvS+oWFhcFVV10VNGvWLEhNTQ2OO+64YNKkSZF6GzduDC666KKgYcOGQePGjYPRo0cHX3zxRQAgVn/Tpk3BmDFjgs6dOwcZGRlBo0aNgl69egXTp0+PvN+CBQuCgQMHBo0aNQrS09ODgoKCYOTIkaFrZMSIEUFGRkb8B0MSrvRaX7hw4SHrTZ48OejYsWOQmpoadO/ePZg7d24wYsSIID8/P1TP9nWl7798+fJQvUcffTTo3LlzUL9+/aB58+bB9ddfH2zdupX+7DvvvDMAEHTq1KnM/VP7q5lK20/pf6mpqUGLFi2Cs846Kxg3blywY8eOUP3yzvHEiRODHj16BA0aNAiysrKC4447LrjtttuCtWvXBkHgG+uff/754Oyzzw7y8vKC1NTUoF27dsHo0aODdevWJecgSJXxzTffBNdee23Qvn37IDU1NcjKygr69OkTjB8/PtizZ08QBEHw/fffB2PHjg06dOgQ1K9fP2jbtm1wxx13xP691DvvvBP07t07aNCgQdCqVavgtttuC+bOnRsACBYsWBCrV1xcHFx22WVBTk5OACDS50r1M3v27ODqq68OOnfuHGRmZgapqalBp06dgp///OdBYWFhrN5LL70UHH/88UF6enrQvn374P777w/+8pe/RMbU/Pz8YNCgQZGf069fv6Bfv36hsn/84x9Bv379gvT09KB169bBvffeGzz55JOR9/S2TzYPkOrJ2y7Lul/Nz88P3W+y+V9ZbTUIDn3vKlVbottOEBy8nx0zZkzQtm3boH79+kGLFi2CM888M5g4cWKszoEDB4Lf/e53QX5+fpCWlhaceOKJwSuvvBLpl8r6juaxxx4LAAS//OUvY2V/+9vfglNPPTXIyMgIMjIygs6dOwdjxowJFi9eHKtzqPtwqX7svQaAID09PejevXswYcKE4MCBA7G6Tz75ZHDUUUcFaWlpQefOnYNJkybR7/V27twZjBkzJmjSpEmQmZkZnH/++cHixYsDAMF///d/H+mPKIdw3nnnBenp6cHOnTvLrDNy5Migfv36waZNmw7rO192X7pp06agS5cuQYsWLYIlS5YEQcDnayUlJcH9998fdO3aNUhLSwsaN24c9OjRIxg7dmywffv2Q36m0j7qo48+Cn784x8H6enpQX5+fvDoo49G6nq/OywqKgr+/d//PWjVqlVQv3794KijjgoefPDB0PURBFVrLE8JAv36TKJ16dIFgwcPpn/BUFFPPfUUrrrqKixcuLDcp2wiIiIiIiIiIiIStmjRIpx44omYPHlyucvIikjFVfvlmKqakpISDBs2LLI+uoiIiIiIiIiIiBxZu3fvjoQYP/LII6hTpw769u1bSXslUrvoIUSCpaam4u67767s3RAREREREREREan1HnjgAXz88cfo378/6tWrh9mzZ2P27Nm47rrr0LZt28rePZFaQQ8hREREREREREREpEY65ZRTMG/ePNx7770oLi5Gu3btcM899+DOO++s7F0TqTWUCSEiIiIiIiIiIiIiIklRp7J3QEREREREREREREREaiY9hBARERERERERERERkaTQQwgREREREREREREREUmKKhNM7Ymm2L9/f6Rs7969oe1t27ZF6nz33XeRstWrV0fKmjZteshtAKhTJ/rcZufOnaHt5cuXR+qkpqZGyjp37hwpa9WqVWi7YcOGkTr169ePlKWkpBxyW3xYO2Rlti2ydjd58uRI2apVqyJlGRkZoe169aKXZYMGDSJle/bsiZRZbN87duwYKRsyZMgh9wkA6tatGylTO6s4do4OHDgQKfv+++9D20VFRZE6y5Yti5R9++23kTJ7frOysiJ12Lm1fd2GDRsidZo1axYpO+GEE8qtl5aWFqmjNpc83jgo2xZtGwCAlStXRso2b94cKbPjWePGjSN12Dm3bZ31t2xcbN++faTMjuue8bSsMkke2z5Zn2jnfwCff82ZMye0zfpOdn7tPvzoRz+K1Dn11FMjZaw/teO62pNUhUi+ePehtveRnuPmuXcAonO7kpKScusAfCz+6KOPQttsXsXuR7ds2RLaZvcJHTp0iJSx8dr+TPbz2H2Op/3UpjYmUlUlc+zyfg9j+0R2X8D6SHv/m52dHanD+iz2/d+R7o/U/1U+2xa97bUqzPfi/a44WfM9/SWEiIiIiIiIiIiIiIgkhR5CiIiIiIiIiIiIiIhIUughhIiIiIiIiIiIiIiIJEXSMyG8a2Dt27cvtL19+/ZIna+++ipS9vbbb4e2v/jii0gdtk76kiVLImU7duwIbbN1fdn6l3YtTbZ2MVtLs1OnTuWWnXTSSZE6bA3iNm3ahLbT09MjdWr7Gq7xYm3YtlfWxsaNGxcpY1kkFlszlq2fatcnZGsYZmZmRsrY+vw9e/YMbbP2ytZDtNSewjzrAu7evTtSZ8WKFZGyhQsXhrbfe++9SB1v/7dp06bQtnf9S7u+JltPn7Wd7t27R8rs+up9+vSJ1OnSpUukzK7fqdwIH886lmzdaZtj89prr0XqfPrpp5GyNWvWRMrsuM7WYmU5SHatV9Y2c3NzI2UFBQWRsrPPPju0zdom2wf7M9XG4uNdK9Wui87yZ+bPnx8p+/Of/xwp++CDD0Lb3rV97T6wvJurrroqUnbZZZdFymxbZDlP6suqp3jzATz14n2dt46njLVB1lZravuNNyuOze02btwYKbPzseLi4kidtWvXRsrmzp0bKZs5c2Zou1GjRpE6bL5n94uNnSNGjIiUHXXUUZGynJyc0HaTJk0idfLy8iJl9p6bzS+ZmtDGRKqCeNeyT+T4xr5DYxmYts9i78XmWTYTbNeuXZE6tg8r673izfqKN/+GfUZlwiaPp33a7wNZHVZWkdyIeM+xp62w+yPPfM/zHWHkZx32K0RERERERERERERERBz0EEJERERERERERERERJJCDyFERERERERERERERCQpkp4JwbD1s+wa6C+99FKkztdffx0ps+sEb968OVKHrffG1va1672x/WRrXjVu3Di0zdaSY2tMb926NVJmsyrY52Hrvp911lmhbba+ekZGRqTMuzZyTeVZn9CzJhxb/5ytCWezO4Doumqs3TE294OdS5YlwTIn7Frt3uNQm9pKPNgxs+v9vvvuu5E6LO/Brn/J1khn57tt27aRMrtOMMsCYPtu1/Fl6w2z/s9mSQDA4sWLQ9s2WwcAvvnmm0iZ7etatmwZqVNT16ZOJNY/sbHSZoqwdsfO7969eyNlts/yrMMPAGlpaaFt1tex19mcJyDaplj7Yf00W0fbUhuLsv0Ia3esrdj8pFdffTVS54033oiUsTXX7XyPjbGecZ7Nod5///1IGetPL7zwwtD2cccdF6nD3j/eNYclOeLNVWDt3vO6eH+e9708axezNmf75LLq1dQsHc/4uW7dukgddq9g723ZvS6777N5TUA0B66wsDBSh/V/tu9hffLUqVMjZZ07d46UHXPMMaFtlu3F1nhv3bp1aJvNJZUTIZIYyR5b4n0d63tsLhx7Lctx9WRvsb6I3dMw9p7Ge+8ZbyZEvK+TxLFjP/tOjd2P2vsC9rp4MyG8bcXzOvY9Eht37T1xPO1VfwkhIiIiIiIiIiIiIiJJoYcQIiIiIiIiIiIiIiKSFHoIISIiIiIiIiIiIiIiSaGHECIiIiIiIiIiIiIikhQJDab2hp6xgJk5c+aEtj///PNInaKionLfi4XJsBBLFlbD6lmewBdvYBsLL7SfkQWEsPd6/fXXQ9tNmzaN1Dn++ONd+yrls8Fu3377baQOC3llwUk2nIa1C9YOPMGHLPiGBaLbsDwWmimHxvo/di6XL18e2v7ggw8iddavXx8psyG7LHSX9X+sXdgQQm8Ykm2HLIRr9+7drn2wfTDrf1lAY4sWLULbLLyQBbzW9rAuT39hQ9OB6BjL+hTWzln7tOfY2+5sf+s9l2xfbQjd9u3bI3VsGwOiwVy1vT152XbG2pgNqQeAN998M7T97rvvRurYvpT9PCDaT3mDqe3rWB/F2vk//vGPSJltL2we3KNHj0hZbm5uaNs7J5SK8/ZPnnsfTwA0K4s3TJqVsf7QW2ax/WLXh+e+qirxHm/Wh9i+gPVrr7zySqTMhk6zORS7V8vOzi53X5s0aRKpw9j5GAuiZPM9Nkf76quvQtuffvpppE6fPn0iZQMHDgxtN2zY0LUPrExqBnY9aryLjycU2ju2xBtMbd+L9aPsuz4WVm2ve28/YOux9sTGQE9YtQ3rLWu/EhlMbd9f10x84g1XZ22Yff9ny9h9s2fuBUTnVd55lud1bOxn36nY18bTzqvX7FBERERERERERERERKoNPYQQEREREREREREREZGk0EMIERERERERERERERFJCj2EEBERERERERERERGRpEhoMDXDAjtYiK8N8PIG09j3Z6Ee3rA0G8bhDaXzhOOwYB/G7n9JSUmkDgsts0G2ixYtitTp1KlTpIwFkNhjU9sDbVg7sO1u48aNkTosQJKFFrFz7Klj2xl7b1bGzufmzZtD2xUJZPT8vJqIHQvWZ61YsSK0vWrVqkgdFkxtQ408IVkAv8Y954T1pZ5gavY61n63bNkS2mb9Gjt+S5cuDW2zEHUbsghUv4DMivBcl2xMYmFa9rywMZ21RW/4r2e/7OtY38qCOz0h2onc99rO0wcuWbIkUmf+/PmRMhvuvG7dunLfuyy2vXjPpSfEjbUxFla9bNmy0DbrO9l+nXLKKaFtFjRbW8bYqiDeuRHr1zyh0N5QZE+gKHudp8wbXu0JM66OIcLsnLM5TWFhYWh7zpw5kTo2tBmI9iuZmZmROuy4sf7IvtZzj8fqsfGUlXn6xE2bNkXqvPfee5Gy3Nzc0HZeXl7c+2Cpj6x64p1TJXMuVlPaiScUmvXh3nHK1vPM24Ho2MLu+9j8iYX42vs8NpYxth7ry+P9zo4F+Mb7PRvr11g9e5zjveeXKM91xNoPa8P2Ow/v99ysHdh7GnZPzOYMtswbTN20adNy98E71wj9+yH/VUREREREREREREREJE56CCEiIiIiIiIiIiIiIkmhhxAiIiIiIiIiIiIiIpIUegghIiIiIiIiIiIiIiJJkdBgam8wqw2hBqLBviysxhuY48ECO2yoBgvCYSEb9r1YAIw3SM4eQ3ZM2We24Zo2BBHg4WAsAK02Bbh6eMKVWMAMC4phZTbUhv08T1v0Bv6xerb9sH1gZQo7+hd2fFg/tnLlytA2C6G2AYeMJ6wI4Ofbc97Y69j7e7A+a8+ePaFtForMxo+1a9eGtrdv3x6pwwIN4z0ONYUnKJUFbNm+Lt6xDIieT2/4m30v73lj7c6+F+tbWZmCqcvHzqcNaLOB0wAPq169enVomwWIM54AV0+4HxANQfX2f+wasX06C1jNzs6OlLVv377cOqmpqa79kuTwhBeyNsH6J9v3VCQ81JZ5g6nt2MzqsLHCBoUC0bZZ1YKpPfdc7HizMcL2Yxs2bIjUYeGNbJ5Y3n4C/LpPT08PbbN7E9ZWsrKyQtvsPHmPg91X1m+yz2y/G+jdu3e5+8l+HlC75nbVgeee0nvfyco8Yzxj6yVynleZbdDzvQXr19n1zO7D4r0vsO/PAnxZGfs8OTk5oW3P93PefWDjGzuf9v3tPgG8n7Z9onffWV/qGWP1vV7i2LbI2gr7XmLFihWh7TVr1kTqsPscdu7s/YOnjbEyVofdm7Rr167ceiyUvTxqlSIiIiIiIiIiIiIikhR6CCEiIiIiIiIiIiIiIkmhhxAiIiIiIiIiIiIiIpIUFcqE8KwxXVxcHClbtWpVpMyuQWq3Ab7mHFurzmLraXnKvOuqxbu+Ojtetoyt3cnWH7Ov27ZtW6QOW2e+TZs2kTK7RlhNWVvTs9ajdz1Ie15YO2Rrv7J122xbj3fdVbYerzc3wLM2p/yLd91gtu6tzb9hWS1bt26NlNn8Frv2L8DbV7xZCJ51duPt14BoP+ZZdxSIHi+2nqc3a6A287Zh24+x9SJZGTufnkwI1mfZet41/Rs2bBgp86zrz9qd+sQwb16V7ctsJg7A+0B7XbO2wvo7T5ti7Yftu+1jvX2pJw+FfWabd8PKOnbsGKnD5ho1Zd5WHcSbCcH6SDsuevsiT04Ea+OefprNP9l9CBuL7byF9b9Vqa1WZG5nr1WWc8Xuie059mZ7eTKPGM/xjjd3hL2W1WGf0Y4V7D62VatWrveSyuPNcfC0E2/WgG0D7HphZZ68MW/mSFXvx+yxZN+zsf6J9XX2td5zZ1/H5kGbN2+OlLF7DE++Gzsnnu/LWBk7prZN5ebmRuqw+xD7OjaHY2XsONh5qjcfQOJj2w+bC7Hr6Lvvvgttsyw8luGUyEwI26Y8dQB+PTRv3jxSdrj0lxAiIiIiIiIiIiIiIpIUegghIiIiIiIiIiIiIiJJoYcQIiIiIiIiIiIiIiKSFHoIISIiIiIiIiIiIiIiSZHQpBIWFsSCyliZxQK9WIiODaLxhkmzffWErsaLheOw97efhwWesAAgG/7GAnpsIG5Z78+Cb2qqeMNGPWFW3lBoT0ARCzZav359aDsnJ8e1D6zdsVAbOTysLbFALxvww/pD1v/Z88aCqePlDWOzvAHQnnAwdqwYG9zkDab2hsvVFt5jZPuGxo0bR+rY8Qfggad27Io33I+9N+tbmzRpEimz142nbQIKpvbwBMmz0EEWembnMN7j75kDsvGOzavse3mDqT1hrZ6xAQBWr14d2mbzYDY/qM19WzLFGwrtDQH1BH562f3yBkzbeux1rL9l7deOF+z+oqq3Ve/czgYrs3HK1gGi/Up2dnakDmtjnntb772JZ2z29sGe9sP60p07d4a2t2/fHqnDxhjWn3vu0SQx7LH29pG2XbBryraJst7fzuvY/RG7t2bXQrzinS8kg+d4s/GH9eEsLNeeK9bXsTL7M1kwNduHjIyMSJkdg9h9s2d+z+akbL/YWGz3gR1T9t2MbYusD2NtuEGDBpGyrKys0Da7R5P4ePoyNr6xcd5+Z7d06dJIHdbuPMHU3mBzW8bqsLkd6wP69u0b2o6nb9NfQoiIiIiIiIiIiIiISFLoIYSIiIiIiIiIiIiIiCSFHkKIiIiIiIiIiIiIiEhS6CGEiIiIiIiIiIiIiIgkRYWCqW1gBwtt8YaGekJWWTCGDf5hISIsHIiV2cArFs7hCSnxBh15AudY4AkL+7FBNGzfWegTC8tjIaMSZs85O94Ma3e2HbAAplNOOSVS9swzzxzyfQ5nH2zYkULcDh/r19g1Z/sxdj2zgC0bIMQC+lj/5AljY23CE87rDWX1vD/r39l15QkvrEiYZ03lCQ9k59MGYDVr1ixSh/VZrH3a9/fug22v7HWsrbB9tSFunnBP9jMVdB7FznlhYWFom80JWZmdm7DgPk8IdSJ5fx5rG54QThZquHHjxtA2C+pUEGFixBs+z15nA5/ZuWXhhXY8Y30RuxY8YzG7n/CUsfBq7/1Ro0aNQtvs/oJ9nsriDdVl8zZ7bbL3Yu3A9iueENay9ive/s+2lXjnhED02LA5Ggsot+2OBdSyuV28160cPnasbRmbB7A+xAYCr1ixIlLHjn8A0Lp160hZ06ZNQ9sswJe1OXtdsc/nvZ+wfRv7Xqsyg6ntnIPNJdi1umXLlnLreb9fsmUsFJpd902aNImU2ePLPg873vYcs/nnhg0bImWsXdv7a9bO2fvbtsjGwIYNG0bK2FzPnmu2nxIfz3yAtXMWMG3vhdasWROpw64/di1brJ2zMtvfeQOtGdvOFEwtIiIiIiIiIiIiIiJVhh5CiIiIiIiIiIiIiIhIUughhIiIiIiIiIiIiIiIJEWFMiEstm4VW0uPrXNm19Jja2KytbLsGpJsbS62TpVnTVXPmtZA9DN612pna7Ha/WLrcrL1uuw613Yb8K/frvU1Dx9bf5LxZJHk5eVF6vTq1StS9te//jW07V23krUDu26vHD523bB1Mm0Z6y88fQ+rw/pgz7qArF16VGT9VLsP3r7IrvvJsiS8fXBt5s00sH0bG3/YetVsbVS7Dqp3/W27zjVrK+ycs/VTs7OzD/ne4sPOHTsHdm1U1ieytXztXI61MXbuPOcz3jwPVsczpgPRddJZG2ZZQHZ9ZFZHfdvh8xwz7/jG5vK2TbP8h7Vr10bK7FrU7Oex+SZrh3a/PGsLs9ex9sz2i92j2X1t2bJlpE5lZkJ4sn7YcWP3mnbMY30RGxft8c3JyYnU8ebR2Pdi58nTDrx9iufYsBwMz89kr2PXmrddS9kqcr5tm2Nr4C9ZsiRS9tZbb4W2//GPf7j2i2UknnzyyeXuJ7tmLTY/Yf0ae6/OnTuHtlu1alXuz0sWz/0om0uwz7p+/fpImZ2XsOPGcnM8eTHsno7Nvey9iDcryWLHgWVVsL7b/kxWh2WD2HshllfCxnn2/pYyEcvnybYpq8z2d+z+hWXZ2OuIjW/smmFjXry5H/Y+irU7VsYyLhLRznT3LSIiIiIiIiIiIiIiSaGHECIiIiIiIiIiIiIikhR6CCEiIiIiIiIiIiIiIkmhhxAiIiIiIiIiIiIiIpIUCQ2mZiFZLGCroKAguiMmLIO9FwtbXrx4cWibBYR4gpSA+MNZLW+YnSfIOzc3N1KHHVMb9tauXbtIndatW0fKWACJJ8intrPHyAarszplsUFD+fn5kTrHHntspMxeM95wNhYsawNc1QYOrSLhhbbMGzDUsGHDcn8e4wlbYsFHiQw79YRrsnbpCfauSDC1LatN7d4bwuUJUPOGadkQt3gDJdl5YmM/m0fYvpqN+95j46lTU9sU+6ysv7MhlSxYj7Uf+14s3M8717LnwHtOPPXYz2P7Za8H73Gw7Zr1d+w6qk1t0Yr3WrVl7Nyy88bCNW2Q37p16yJ1WJm9XlibYDz3E6wOaxO2rbLrmo2x2dnZkTI7T2Vz2YyMjEhZZfH2Key82LbhDUq17+8Jti/r/e059s4v7T6w/fQeG891ZNsFwwJjvZ8nkfPXmiiRgexr1qwJbf/973+P1LEh1ED0+xsbdgzwORybg9rvRdj3TOx6sdfV2rVrI3W++eabSNnq1asjZRdddFFo234vcyR5zh27vrZu3RopY6G0tp43RN72kWw+w8pYO7B9FLuHZK+zWJtmx4HdY9ixyxNeDUTnsyyEmo2LrF9O1H2VRLFjab+D8IRQszI2prOxko15tl684x0bh9k+sLZvr5t49kF/CSEiIiIiIiIiIiIiIkmhhxAiIiIiIiIiIiIiIpIUegghIiIiIiIiIiIiIiJJoYcQIiIiIiIiIiIiIiKSFBUKprYBKSzgkQUrn3TSSZEyG8a7cuXKSJ1ly5ZFymw9FmjDeALavMFcHuzYeMJTWbDRCSecECmzQdStWrWK1MnLy4uU2bBboPaEFyYSCxBiIUmegPI2bdpE6jRt2rTcn+kNomT7Zd9LbeDwsT6FhfnYMCLWJtg5ysrKCm2zUDG2D57ASna+WaCXfS9P6FdZ+2DD5VgwF+vP7edmxzjZQdtVifdzeYIIGRtcxdorCxQ80lhbYWF5tt15Q31ravuJFzseLJTPngNPoGtZ72+xtsjY+Vcirxnve9k+1hs+6+nvaksQYbznjR0fdvxtGTsfrK9jwYSFhYXl1vGETrP9tMGIZb3Xli1byq3Dxn57vNhcg93TNGnSJFJm55YsBLQ68lxzLPTRe18Q7z54QqHZvC3eoGJWz35ub0C3LfP+PA/v/VF1E+/xsO2C9Q2sz1q0aFGk7N133w1tL1++PFJn1apVkTIb/svGNjY3YMHXtt6xxx4bqcP6J/szbcg2AHz33XeRMhuqDUTDsU877bRIHU8geyKwa8f242wcYePb5s2bI2V2bPEE17Kf6T3nnu/Q2LH19HVsjLWfD4j/fpQFqdv7XXb/y44D+4z2tbpXiY93nrh9+/bQ9ooVKyJ1WHD9jh07Qtue7wOB+M9nIsc31hbtcVAwtYiIiIiIiIiIiIiIVBl6CCEiIiIiIiIiIiIiIkmhhxAiIiIiIiIiIiIiIpIUegghIiIiIiIiIiIiIiJJkdCEHBaCwYKPW7RoESlr1KhRuXXYe82dOze07QlhLWtf4w3uTGQooN1XFibdq1evSFmzZs1C2ywomQXfsLCfmhDWlWz2GHmDqVnIjQ0aYm3fhhIDQHZ2dmibBYuxUCZPSJLawKGxvoH1MywY154nFqKcnp4eKbN9JAsVYzz7yvoBT7/m7VsZ2583btw4UseGewLR48eOQ20KpmY8Abre42FDJTds2BCpw8L8PMHmjOd1rL2yELy1a9dGyo477rhy34vxHL+a2m96A5k9wcos4IyNi/a8sEC+eK/peMPfKhIaZ8ddG1IH8ONny1g4Yk0Ipk7kubT9P+sbWICkHUts8B4QDVMtq549b6xfY3NEu++eOmXtl+2r2XHw3AOwazYzMzNSxq5j1l5rKjtvY/es7LjZ4x1vALsXu2YSOXbZvprdv7Aye/xq03iaqNexOTkbVzZt2hTa/vzzzyN1Pvroo0gZm+vZ/s++N8D7AdsH27kCwPssFp68bt260PZbb70VqcOCqe0xZfvA+lv2eT7++OPQNjvu7P47GVg7sMeSHVs2L/GEVbM67FjaY8KOEWvnbGyx99dsLGPjrsXGN3Yc2L2mbcNsnsrOuf3OyH6fA/Dj4PluoDbd63p57t+83+HYoPolS5ZE6qxcuTJSxtqZ5b3H8JzjRLYDdp3a+2t2/Nj89Yf0lxAiIiIiIiIiIiIiIpIUegghIiIiIiIiIiIiIiJJoYcQIiIiIiIiIiIiIiKSFEnPhPCuSW/rsbXdWD6CXX+tImtp2rW4PLkRrF4i1+FiOQ5NmzaNlNn15Ng6XGytOs96eRJlzzlb+5WdO7YmnF3jj62Nz66ZnJyc0DZbP5+dX7amYDLXqawta6eztfw8a5yydfRYxohd99GugXo4+2XbRTLXqAV8awWy9VoZew2xdTrZdaZ1MsPY8WDjri1j/QxbE529V7y5FJZ33GLXiG2LyV6jt6b2f6xfYeuG2muRre3rWfuWzV9Yv8Lana2X7DXX2fm1n4ftJ1t72paxNWrZcajK7S6RmTXsmG3cuDG0bdevBvi4YcvYmtneLCbbR7E+K5Hng12Pdr1qNi6yuas99qyPZGv62/wg9jNryj0HO962j2LHw87bgegxqUj2TKLGWG/ODKtnPzc7Duz+yNZj712RY1NV2M/A+m9vvqUdc1kewzfffBMp++yzz0Lbq1evjtRhYw0r27JlyyG3Ad5vxttWWT37/iy7gmWE2XGYfXfimVMAwOLFi0PbLFeArf2fDJ6sJDYeePOTbAaRNxPC7oO3n2HzRk8egydjjr03m1d4cjfZz/OMi6xNs7GZ7UNNyARLJM/80pudw/pFm/2ydOnSSB12T2zPk3d8ixebX3rmnOz4sb7C5l6wY1qemjEbFBERERERERERERGRKkcPIUREREREREREREREJCn0EEJERERERERERERERJJCDyFERERERERERERERCQpkh5M7Q2rtoEyLAiNhbuwwCCLBX3EG87BeIKpvcfGlrGQJE8ZOy5VJZSwJmLtlZWxoCYbPsSCq1jbt0G+69evj9TxBlPb9qO2EuYJUGNBVux8sxAsi4WT2/PGQoDYeWP76gk/SmQbYPtq26YN3mZ12HuxIDUFU5ePnV9PgBoLpmaBXoyn3XnqeIMcbXgeEN1X1t/WlPDUZGLXE7sWPWGITEZGRmibnRP2XvH2W555W0UCoG3/zeZx7PPY48fGlHgC4SqTJzTT+zoW/rlo0aLQNgtT9YZRWmxe5wkuZW3CE6TJzi0LAWXzP1bPsw+JDGu17bU6zi1Zu2PtxxNSavs1IHoObLB6WTxzmnjnPd5+jZ1zGzrNAlbZsfG0fXbcPUGklYXtrw2PZiGmLMiZXc82bHn58uWROmzOZn+m9+exfbXXuDc8154jz3dDgL8dlvfzgGg/5r1fYvtg55tsbDpSPNeOt19j46JtL2zu55kPevsZz76ycSreYOp4+x728zyfkc0rvPtV3YKpE9k3e/t+e9xYH7Vu3bpI2cKFCyNlX331VbmvY9eM53skJt7jFW8wtfd7JBva7Zk/W7rTFhERERERERERERGRpNBDCBERERERERERERERSQo9hBARERERERERERERkaTQQwgREREREREREREREUmKhAZTM/EGMrNwF29YhuUJRGLv7wm99u4Dw97fE1zCXucJRIs3pETKxwJdWUgSCxXKysoKbbPgOnbObTC1N1DLG6xoqa38iyf4COBhb57AShaWa9sT+3ksxM0TCs3EG87r7adtGWv3nmuIBUx5g6ltWW1q4+z8suNtjxELe/aOsbaMvY614XiD11iIb1FRUWi7VatWrn2oTW3Dg50TTxCht9+y/YE39Mwzt0t2IBw7NjaY2m4DvlBINqZ4QvCqEtY/237FEygJAKtWrYqUffrpp+Xugyccm83r2Hnzzv8sdr7tsWHhsCwI8euvv46U2QBctk9s/MzMzAxts/kha3PsmDZq1Ci07Q2mryzeeTQ7brafYWOGDW0GonPyDRs2uPYh3jGJvZftlytyD25fa9uT9/09+1nV2fkGAPzf//1faJsFR7P2VVxcXG7Zzp07I3XYmGEDptm8js2f2Bhv99Ubbmz7Au/59vQ93jmj3VfvvJj1ifY4sPN1pLDr0n5Wdmy9gcz2HHjOL3t/9vPYsWXn09ZjdTxB0fF+ZsY7Vto5rzck3BvEXJXEe23Gew/Jymxb3LhxY6QOmzcuWrQoUmYDmVk/6bkn9p43zzjP2jlri7bdeb6HLqts7dq1oW02ZpVHfwkhIiIiIiIiIiIiIiJJoYcQIiIiIiIiIiIiIiKSFHoIISIiIiIiIiIiIiIiSVFlMyG8a0/Gu76oZ33NZK+J7lkXzfv5PMdPa1onjl1XjeUssPbD1h60aww3bNgwUoet22bX2mW8mSKJahue9bhrAs96mwBfI89e96yd5OTkRMrscWQ/j61N7el72DlibTXetQk9612ytZLZWvGe9Tyr2xrplcG7hqQ9vmz9y0Qe23jX9PfOD+w6vd7jEK+a0P955y/xrgHM+i3bH7C1qBOZ7eD5jN512T19J1snna3/b9/Lk7sBVO3+jvUhNtvBm1/wz3/+M1L27bffhra9+QX2WLP8B09eExAdu7zzQbuGvF13FwCWLVsWKVu/fn2kzK4F37hx40gd1tfZ88PWtWfHlJ3X5s2bh7ZZ+63qvGOL556R3StY3jWzPXlficxr8o5ldh/YPQ2bv3rGee866VWFzWUBgCVLloS2PTkLZdWzGRAsE4Jdvzt27Ahts2vXm7dm67G+wbPuvjfvw5Nd5skQAKJzD3bPwcYBlmFnj40nFyhZPN9DsePtzTu0Zey69Jxz79zSk+2QyEwIb1l5713WPnjmxfHm31Rmf8h+tic/xDtO2Xqs3XkycGwfDAALFy6MlLHMMdt3eu57AN95Ydetpw1772MTeW+7efPm0LY3t++H9JcQIiIiIiIiIiIiIiKSFHoIISIiIiIiIiIiIiIiSaGHECIiIiIiIiIiIiIikhR6CCEiIiIiIiIiIiIiIkmR0GDqigQweoKVWdCHDURir/OGfNnXxhvMyrBAEhaAZH8mC1hhPPvl3feaEKSZbPbcsXPJjqMnVI0F9LLQGRYyaLFgsaoc4ladecKXWT12blmQn31dRQK9rMpoEzbMiQU2sjLbJ7Jj7A0vrM3XgjcAyxOM6+Vpw54wtnjDz4FoACM7Dt6y2oydO9Y2POecBT/awEhPYHlZ9crbp7LY94+3f2X7xYItWTC1fR0LwWPt3LMPldWm2byksLAwtM3mVOz4LF++PFJmA/NYWCubW9vzy8bhnJycSBkbp2wAIGsn7HrZsGFDaHv16tWROlu3bo2UsXD3Zs2ahbbz8vLK3U8gei5sECPAjx/7PDaAPZ7wwmTy9BeegFXAF/rIzpPdh4rMSzyBsZ5+M5HzJXZ9eEK1472HqkpYP5aVlRXa9o5jnjbH2pf9eex1bDxi/TQr88zZWEiz5z7ae19gPw97L1Zm98tTB+Bjg+3b2HE/UjzzkoqEz3vCnT1h1d5AYta32vdi585zrxBvKDIr8+wne3/vnJSxn9E7J00G9lmLiopC22w+4Q3mtm2K9UfFxcWRsrVr14a2//nPf0bqeOdadp7D5uRsnuMJZWc896Os7bOxwL6uIuO8vZeOh/4SQkREREREREREREREkkIPIUREREREREREREREJCn0EEJERERERERERERERJJCDyFERERERERERERERCQpEhpMzSQy9JGFf7BQEssbsuEJjK3I+3teZ8tY+JsnBEoh1Mll2wYLrmLniYUP2QAbb8i1Dfxj782uD28Ik/yLJ7SPHVdPIBU73yz0zIYAsX3wBvbaehUJXPVg7+UJpmZhebZPjPe4S5QnCNIz/gC+fibeUCxvkKOnT/ReR7VJvP0dG28855P1d54+iqlIwJ+VyMBYT+gxCzW0r/OEPQJVu79jfb0N8fQGHzN2buQNs7X9CjtHnoBVVsb6FHbe7JjXunXrSJ127dpFylq1ahUpa9q0aWibta8tW7ZEyuznZuGMNmwS4Nd/t27dQttsTD9SPNeEd0xKZDC1nQtVZG4X73Xv6Te9fasnMDbee0/v/KOqaNKkSaTspJNOCm2vWbMmUoeFt7LA1Z07d4a2Wb/pCUn1jiHsvNm2ya4DT3Aqex1r9/F+h8TEu+9sDLPnojK/X/EcN+85Z2WegGlWZtudt0/xtkVPHU+odkUCey3vsbE81xrA5ySVhY2Lq1atCm2vX78+UsfTxliZveYAPl+xP3PlypWROhs2bIiUsT7XfhfjDaa2+84+n7fPsOec9UdsrmHPD9vPeL+398x/rNp9py0iIiIiIiIiIiIiIkmjhxAiIiIiIiIiIiIiIpIUegghIiIiIiIiIiIiIiJJoYcQIiIiIiIiIiIiIiKSFEkPpk4kFhCSyGDqRAX5eYNFPKFi7DOz4BcFTB9ZnmBqb9AOO58eDRo0KLcOC8ypyoGV1Vm84VMsPIgFYrIgSMsbZHWksf2yfTcLf2Mhlps2bQpte0Nza3u7t+fAG/hnj2VFxlNP0K8nSM57zhlPsCUL2NIYG8bOAQs5s/XYcfSGNMfrSF/7njbMPjPrAz1hdt7g5aqCjXm5ubmhbRY4yK5xNke2ZWwcYfcOts2xsD9vWLXF2gS7Xho3bnzIfQKigdMAD7C2Yd/sM69bty5S1qhRo9A2C8n1BoefeOKJoe2cnJxInarOey3ZMZW9jrV9e14qEtYaTx2G9TPegGnbH3kDhz3v7Z3bVZX+j/UhxxxzTGibhcqz+b7nOxAbmgrw69LeG3qPq6dtxnvP4Rk3D6dePNi+e4OprczMzITsUzzY5/Dss5dtB/G2H++1y97f0996AnW99+6ee4B4+yJ2vthY4Ql4r8x7Ffad0+LFi0PbNqgaiD/YnM1p2Hxl8+bNoW0WQm3rALzPtf0pm8d5+knvuMiuW9u3eL4PBKKfh+0na2Nsjmt/ZjztrvK/nRIRERERERERERERkRpJDyFERERERERERERERCQp9BBCRERERERERERERESSolIyIeJdV42tzeVZ3zmR64Mncq01z9pxbN1Vtv6Y5VkHT+Jnj6V37T62drNnzWeGvX957w0ktx3UljbmXffR08+wtfbYGta27XjzH7zrC3t41ndnWD27diTb9+zsbNd7WfGuqexdU7Q68mRCsD4l3rEy3vWq482E8O6n7RPZcfDmZdRm7BrzZEJ4r3O7XmpFMiLsPnj7B3vOvfMq9hnt/rNsAVbm2ffqloHDzqUd87xzKsYe/23btkXqsHXS7blk54OtvcvW7LXvFW/f6s1/sFkSbB9YzgZ7L5vPwY4fux9j13+vXr1C24lcl7wysevetmtWh7VrTy6ct89KZIZTPHWA6Odhbd+TL5HIXKDKwj6nzWph/Qy77tl66/aaY2uks+vSnqNkf0/imVNVJL8uUXN5z3UN8Ps2W2azdY4k9jls3+u9LlmfZT+rNy/GnmPvdxSesni/96rIfN/TZ7G2Yo8pGxfT09MjZayvsPPnysyBZH2UzXHcsmVLpA77rpO9l+3f2DyO5eJs37693H3w5g7Hmytoz7E3D7R58+aRMjvfY/0dm+/Fu+9s3tukSZNy36s8+ksIERERERERERERERFJCj2EEBERERERERERERGRpNBDCBERERERERERERERSQo9hBARERERERERERERkaRIejB1vOEu3mDqeAO9PKFF3sAcTwhMvCFJnmAWJt5gn5riSIcyegNmPIGq3gA6G3bE2iH7ed4AKTk83rDceMMvbbtg5zHZgVSeIDAWauQJG2avy8nJiZTZz5jIkPDqKN4x1hOeC8QfWJnI19l9r0iIon2ttw3b41Wb+kx2bOM9Byxsj4Vw2jBNNsYy8e4Xe51tB6zPZaGDjA2ls+GkAO/v7D6weal3zlCV2euLXYOs7bBAZjtHZuMpC2u1bFh2WWWe4E6G7YNtF82aNYvUYYGA7NhYLACefR77Xjt27IjUsUGPAL83admyZWi7OoYNs7GS9QX2s7HXsfNk72PZ61gZ6wvinQMmMrTXM8Z6wtw943BZ719VsP21/QXrv1n/wY6ZJ5DZMz5450+euXy8c9KqKt7vU1gfcaR4wmXZvad3zLNjCQsIZu3OjnnsOzzWFj3Xvbe/KO99DofnnLPr1o7zeXl5kTpsXpybm1tuvao2xtp2xuZj7LtO1jZsPTbnYCHXtoy1Te+5s32zd3yz1wyb29n5EgC0bt06UmZ/5vr16yN1tm7dGimzx6+oqChSh81n2f0K29fDpb+EEBERERERERERERGRpNBDCBERERERERERERERSQo9hBARERERERERERERkaTQQwgREREREREREREREUmKhKbmeMNd4g2eZGEjiQwb9QQ8xRvKFG9YWLzB1JJc9hyzIBcW5sTYEKF4g6m9oUxsX6tLQFh14wn/ZUGFLLjJ8zoWhsTak6eNeYLA4g31ZfW815AnFKqmhlAnkjdsz9NW4j0H3pBd21Yqcs7jHefVR4Z5Q8Vt2BsLx2UBbTYM1xO8W9Y+2DblPee232L9mDeY2u4/Cxdmx8EGPrJ98M4Z4g0QTTTPeMDOETvWbIywQY1sbPEEU7NxmJWx/fLc07DXNWrUKLTNwsozMzMjZeyY2mPIghfZcbb9MptXsFBTdkzt54n3XuhI8d6/sTZljy873ux1NoCTzaE8AdBens/oPQ6snv08rA5rP545ITsOVXm89rR37zXhOd+sTlWdD1eVc/RDiQzVrsyAYNambF/MwpBZmC0rs309m5/ZORwQnc94xmGAH0vbv7LxlPXB9txVJOze810QmzPYeTA7F23bto2UtWvXLlJm542VGYjOfrbdZ3a8N23aFCljY8SuXbsOuV3W6+zxZt+jsu9bPX2n5/wCQOPGjUPbLHCanV8WAL1t27bQ9pIlSyJ1CgsLI2V2vswC5dk5ZJ+xTZs2oW3W9stTtWeDIiIiIiIiIiIiIiJSbekhhIiIiIiIiIiIiIiIJIUeQoiIiIiIiIiIiIiISFJUaOEwuxaad91Bz1prbK1otnZVvKrLGolsvTzPumXeNTKr4pqMiXCk14lna+aytYo965l61/S066+x9RDZ2nts7caa2g6SxXu8POtYZmVlReqwMrsONFs/nJ1bz5rPXp51g9l6gqwd2jK2jiM7DrbdV/U1pqsKT5v15IdUJPspXp4cB7bvnvXz470WxMf2B3ZdYoD3ZfbcsbXxvecu3nNs27p3jGVldj7APvPmzZsjZXZdZW8mRFWd4wK+cYPtPxtP2TmxYyV7nSePho1l7Od5cpA8bQKIjnmsTrxrPrPjzuYM9md6MjwAfr/C1hKuSjw5V95MCHtM2H0BK/OsK+5Z2xyItmtvRpcnz4L9PE8mTrz3Id5sjOp2/+Jpc4wn98WbQ5DM8aG6nY94eD5jZR4H1g7s2GKzkwC+Jv3OnTvL/Xksp4it12/X4vfOzTz30onMi4n3OzTWb3q+H2revHmkjjcfwGYNVGYWCRu7bHYA6/tZRhrLIrFtkbUx9l2x/d6UfY/K5tGe+QCb49hzAkQzPti5ZPcFrP0UFRWFttl3JevXry+3jGVjsPbDvp+x7ZPtZ3n07Y2IiIiIiIiIiIiIiCSFHkKIiIiIiIiIiIiIiEhS6CGEiIiIiIiIiIiIiIgkhR5CiIiIiIiIiIiIiIhIUlQomNoj3oBgT0AmEA2BYcFoyRZv+FC8gTn79u2LlFXlEMLagIURsaAYFuJmX+sNmbTv5Q1MZOExnnBftg+1IYDMK95AQxbUykKNbIiRp02UVY+1MSve8C5vEKKtx4KpsrOzI2U2ICmRYWQ1Wbzhzp6gy6p6bD0Bbd7+VmNs+TwBpKxOixYtImV2XNy0aVOkDgs19Jw777m07dobQs3anQ3WZkGE7DPa+V68ocRVCTtmNtzPO46wMjsOsjkPm0fb9uQJ0C5rX20ZG5tZ2J8N+GRBj575GuP9PCw40rMP7JjaY19Vx4pD8QZTewIrWYCrPd4sjJydE3a8bRmr47kv8F5/nhBtdv2xuZ0NyfTcLwHVb74X774l8h6sKh+fRKjpn688njGWXYMsrNoG6gLRsZKNZcXFxZEy+x0d658YzzyL1fHel8e7D/b9Wb/J+m47DrA5cKtWrSJl7LsBO17E+/kSgfXPubm5oW02HrDPxQLRbeg0C1bevXt3pMwGUbPvir39q/2MbExn32fYMntPAPC24rmv8h5T26ZYiLc3XN2GaMdzb6K/hBARERERERERERERkaTQQwgREREREREREREREUkKPYQQEREREREREREREZGk0EMIERERERERERERERFJikpJuPMEBnnDIm0IjDdYxBPK6Q25sWXekCxPwBYL5vIcm4qEMtX2QCcPTxgRCxxmQXW2zNv2bVAMC8dhWOidzvmhea5xdq2yMB9bxgLMbZATABx11FGhbW+okSds2CveoFbP8WrdunWkzrp16yJl9nixfpSdi+oWXphsrE9h7cK2H28wazLD31gdtl+eMdbzmSXKE3wIRPskVocFH9r+jQXQsVBDT7uON+CTtSdvMLUNicvPz4/U2bhxY6TMhjuy4+ft76oKT8CtJ3CavQ6Itgt2zOIdF71Bl7aMzQXYfnkCgr3n1nNfwK4Fu1/suLO5Bjt+VSmY2vP5vefXE+jN5tpsbmcDK1k/wOaJnr6Otbt4g6nZZ/ZcM3l5eZEyFmS7bdu20Da7PipyPVR3teVzSsV57rlYH87Cqtn1a4N+2TjMrl8bCOz9vsNz/xbv/ai3z/fsA+uf2HGwx5nd/7JwYzYO2Pdn+3CksDHCjoOsrbD5PSuz7YcFTLMxyc4JWZ14g6nZGMvGN/sdHTsO3nZnzzF7L9ZW7H0Iu4fyfr9lz6uCqUVEREREREREREREpMrQQwgREREREREREREREUkKPYQQEREREREREREREZGk0EMIERERERERERERERFJioQGU3tDzzzY61gAyZ49e0LbRUVF5dYB4g+gizfclNVhwSW2bNeuXZE6nn1nxy/eoFCJ8rRrFg7EXrdz587Qtg2iBPg5t+169+7drn3wBLHGe93WFp6gIICHBdnwHhYOxkLNu3fvHtq27aasffAEf3nPdyIDveyxYSGOLFjJHj9vMKyn767JwX+ez+o5n6ytsDF2+/btkTJP4HO84W/svVmf6FGT20E8vNc5C2iz/Rurw8JabVAfCz1j8z1vUJ3lmb96w5JZEGGLFi0OuQ0AzZo1i5StXbs2tM3GFCbeueqR4Amm9l7j7HjYeux1nrJ4A8yB6P6z9svajq2XyHPmnbd46rB9Z8crnrDCIyneOQ27f7MBruy4dejQIVJmx08WyGlDm4FoOCwQ/xhrX+epA/B5W/v27UPb3bp1i9Rh+26vZRuiCfD2VNvndiKWp69n9542PBfg16otY30D6zftnNwzNwN4X2r7AjYmeV7nHRc9YwM7pqyPtPf4TZs2jdTJycmJlLFxx86pK/O7Ps99ATtP7L4g3oBpT5lnnCyLbRvs87AyO755vyth7Gu99yb2OLPj4J0D2fePp93pW2kREREREREREREREUkKPYQQEREREREREREREZGk0EMIERERERERERERERFJipSgEhZ/Zz/SlrE16OzauACwaNGi0HZhYWGkDls3mK27b9eqY2sLe9Y8ZetwsXXc2Hpvdl24Nm3aROp06dKl3PfyrsNa29fJ9DR/tmaaXV+OrYn+1VdfRcpmz54dKWvSpEloe+jQoZE6bM3s7777LrT9v//7v5E6dv08ABgyZEik7Jhjjglts/X54l2/ria2Me/68/YcAcDixYtD282bN4/UYde47Vc8WQ+HUxaPiuTmeNoF6/M//vjj0LZdhxngx4+tuZnM9berOtYGWH9h2/W6desidb788stI2YoVKyJldtxla1971qNk/RO7jmy/BgAFBQWhbbbuNFtn3u5DbW8rbI62ZcuWSNmaNWtC2+zYsnXSbX/Hfh5rP6zMzuW868Hac+xd19+TjcHWHF6/fn2kbNWqVaFtO18AgLZt20bK2DrE1T0XzDtueTKPEpmVxNhjHW9OUUV49j/e+YH32HjWLq5K2Odia0yzvD6W02Wx69K2RXbPunXr1kgZy4mwZey9GNs/edYxB3iOja3H1klncw17/Lxr1nvvV0RqM9u3sTGQXZee+wLW97F7Yjs/897HenKJvJkQnhykeO9jPRkUQHQe7MmPBHi/VpvuTTxzNM98ryLfi9jj6zkn7HUVOU+JOg5ens+jTAgREREREREREREREaky9BBCRERERERERERERESSQg8hREREREREREREREQkKfQQQkREREREREREREREkqJSgqlFRERERERERERERKTm019CiIiIiIiIiIiIiIhIUughhIiIiIiIiIiIiIiIJIUeQoiIiIiIiIiIiIiISFLoIYSIiIiIiIiIiIiIiCSFHkKIiIiIiIiIiIiIiEhS6CGEiIiIiIiIiIiIiIgkhR5CiIiIiIiIiIiIiIhIUughhIiIiIiIiIiIiIiIJIUeQoiIiIiIiIiIiIiISFL8f/UlP5KoYnm2AAAAAElFTkSuQmCC\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "fig, axs = plt.subplots(1, 10, figsize=(20,10))\n",
        "axs = axs.ravel()\n",
        "\n",
        "for i in range(10):\n",
        "    axs[i].imshow(mean_items[i])\n",
        "    axs[i].set_title(f'{classes[i]}')\n",
        "    axs[i].axis('off')\n",
        "    \n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 60,
      "id": "4a61e5f3-4450-4a98-a9a4-c7e1ee9315a3",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:10.433327Z",
          "iopub.status.busy": "2023-05-04T02:16:10.433044Z",
          "iopub.status.idle": "2023-05-04T02:16:10.438635Z",
          "shell.execute_reply": "2023-05-04T02:16:10.437364Z",
          "shell.execute_reply.started": "2023-05-04T02:16:10.433299Z"
        },
        "id": "4a61e5f3-4450-4a98-a9a4-c7e1ee9315a3"
      },
      "outputs": [],
      "source": [
        "def mnist_distance(a, b): return (a - b).abs().mean((-1, -2))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 61,
      "id": "23c12d69-4dcf-4ce3-9127-98d1ed195b10",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:10.441663Z",
          "iopub.status.busy": "2023-05-04T02:16:10.441418Z",
          "iopub.status.idle": "2023-05-04T02:16:10.526844Z",
          "shell.execute_reply": "2023-05-04T02:16:10.526009Z",
          "shell.execute_reply.started": "2023-05-04T02:16:10.441639Z"
        },
        "id": "23c12d69-4dcf-4ce3-9127-98d1ed195b10"
      },
      "outputs": [],
      "source": [
        "test_data = datasets.FashionMNIST(\n",
        "    root=\"data\",\n",
        "    train=False,\n",
        "    download=True,\n",
        "    transform=ToTensor()\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 62,
      "id": "9f6a9ccb-d6f5-47f9-adad-5c3e53b2598d",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:10.533417Z",
          "iopub.status.busy": "2023-05-04T02:16:10.533156Z",
          "iopub.status.idle": "2023-05-04T02:16:12.835373Z",
          "shell.execute_reply": "2023-05-04T02:16:12.834567Z",
          "shell.execute_reply.started": "2023-05-04T02:16:10.533393Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9f6a9ccb-d6f5-47f9-adad-5c3e53b2598d",
        "outputId": "002eda8f-c7ba-4387-852f-186da04b922f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "All good!\n"
          ]
        }
      ],
      "source": [
        "test_data_by_label = [[] for i in range(10)]\n",
        "\n",
        "for x, y in test_data:\n",
        "    test_data_by_label[y].append(x)\n",
        "\n",
        "for i in range(len(test_data_by_label)):\n",
        "    if len(test_data_by_label[i]) != 1000:\n",
        "        print('Error. Unexpected length.')\n",
        "print('All good!')"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "ed4c76d3-f62a-47ed-9cf6-e9fbbbcb43c3",
      "metadata": {
        "id": "ed4c76d3-f62a-47ed-9cf6-e9fbbbcb43c3"
      },
      "source": [
        "# Samples from a category"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 63,
      "id": "3e491ef2-7b68-493d-abea-73fc00f8b22a",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:12.837108Z",
          "iopub.status.busy": "2023-05-04T02:16:12.836624Z",
          "iopub.status.idle": "2023-05-04T02:16:13.552920Z",
          "shell.execute_reply": "2023-05-04T02:16:13.552161Z",
          "shell.execute_reply.started": "2023-05-04T02:16:12.837080Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 138
        },
        "id": "3e491ef2-7b68-493d-abea-73fc00f8b22a",
        "outputId": "8212680f-7b62-4269-df98-43f96d6b4993"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 2000x1000 with 10 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "category_index = 7 # 7 is the label for sneakers\n",
        "first_images = test_data_by_label[category_index][0:10]\n",
        "\n",
        "fig, axs = plt.subplots(1, 10, figsize=(20,10))\n",
        "axs = axs.ravel()\n",
        "\n",
        "for i in range(10):\n",
        "    axs[i].imshow(first_images[i].squeeze())\n",
        "    axs[i].axis('off')\n",
        "    \n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "3a004021-f23f-407f-8a88-6a00abab905b",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-03-28T19:22:28.612224Z",
          "iopub.status.busy": "2023-03-28T19:22:28.611316Z",
          "iopub.status.idle": "2023-03-28T19:22:28.616754Z",
          "shell.execute_reply": "2023-03-28T19:22:28.615919Z",
          "shell.execute_reply.started": "2023-03-28T19:22:28.612196Z"
        },
        "id": "3a004021-f23f-407f-8a88-6a00abab905b"
      },
      "source": [
        "## Try it yourself:\n",
        "Change `category_index` to see samples from a different category"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 64,
      "id": "8e9a4500-e237-43f0-bb8a-fa72244ea3b9",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:13.554847Z",
          "iopub.status.busy": "2023-05-04T02:16:13.554294Z",
          "iopub.status.idle": "2023-05-04T02:16:13.819277Z",
          "shell.execute_reply": "2023-05-04T02:16:13.818250Z",
          "shell.execute_reply.started": "2023-05-04T02:16:13.554818Z"
        },
        "id": "8e9a4500-e237-43f0-bb8a-fa72244ea3b9"
      },
      "outputs": [],
      "source": [
        "all_distances = []\n",
        "\n",
        "# loop through categories\n",
        "for i in range(10):\n",
        "    category_distances = []\n",
        "    category_test_images = torch.stack(test_data_by_label[i])\n",
        "    \n",
        "    # loop through means for each category\n",
        "    for j in range(10):\n",
        "        distance = mnist_distance(category_test_images, mean_items[j])\n",
        "        category_distances.append(distance)\n",
        "    \n",
        "    # all images of a given category were tested against all mean items\n",
        "    all_distances.append(torch.stack(category_distances).transpose(0,1))\n",
        "    \n",
        "all_distances = torch.stack(all_distances)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 65,
      "id": "690dff9d-cc7b-4e91-93a9-e409551c3084",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:13.823549Z",
          "iopub.status.busy": "2023-05-04T02:16:13.822935Z",
          "iopub.status.idle": "2023-05-04T02:16:14.135810Z",
          "shell.execute_reply": "2023-05-04T02:16:14.134572Z",
          "shell.execute_reply.started": "2023-05-04T02:16:13.823521Z"
        },
        "id": "690dff9d-cc7b-4e91-93a9-e409551c3084"
      },
      "outputs": [],
      "source": [
        "accuracy_by_label = []\n",
        "\n",
        "for category in range(10):\n",
        "    accuracy = [1 if torch.argmin(all_distances[category][image]) == category else 0 for image in range(1000)]\n",
        "    accuracy_by_label.append(np.mean(accuracy))"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "97515bd1-83b7-4dae-98b8-af726e4e6998",
      "metadata": {
        "id": "97515bd1-83b7-4dae-98b8-af726e4e6998"
      },
      "source": [
        "# Accuracy breakdown by category"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 66,
      "id": "8a7db4df-685d-44c6-b34d-f33dcf9c7ce5",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:14.141194Z",
          "iopub.status.busy": "2023-05-04T02:16:14.140804Z",
          "iopub.status.idle": "2023-05-04T02:16:14.160144Z",
          "shell.execute_reply": "2023-05-04T02:16:14.159230Z",
          "shell.execute_reply.started": "2023-05-04T02:16:14.141149Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8a7db4df-685d-44c6-b34d-f33dcf9c7ce5",
        "outputId": "bb43eba9-0b6c-4c6d-e76a-45f38683aa88"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "T-shirt/top: 0.623\n",
            "Trouser: 0.933\n",
            "Pullover: 0.254\n",
            "Dress: 0.702\n",
            "Coat: 0.656\n",
            "Sandal: 0.529\n",
            "Shirt: 0.18\n",
            "Sneaker: 0.918\n",
            "Bag: 0.513\n",
            "Ankle boot: 0.816\n"
          ]
        }
      ],
      "source": [
        "for i in range(10):\n",
        "    print(f'{classes[i]}: {accuracy_by_label[i]}')"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "c6c11e80-6803-4afb-aeeb-1e590319e395",
      "metadata": {
        "id": "c6c11e80-6803-4afb-aeeb-1e590319e395"
      },
      "source": [
        "# Average accuracy of the baseline"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 67,
      "id": "499b8d1a-bcad-4a84-9338-de0262310ebc",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:14.161726Z",
          "iopub.status.busy": "2023-05-04T02:16:14.161286Z",
          "iopub.status.idle": "2023-05-04T02:16:14.176358Z",
          "shell.execute_reply": "2023-05-04T02:16:14.175579Z",
          "shell.execute_reply.started": "2023-05-04T02:16:14.161684Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "499b8d1a-bcad-4a84-9338-de0262310ebc",
        "outputId": "8814b88b-d558-4986-9683-510ce82be6a1"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6123999999999999"
            ]
          },
          "metadata": {},
          "execution_count": 67
        }
      ],
      "source": [
        "baseline_accuracy = np.mean(accuracy_by_label)\n",
        "baseline_accuracy"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Looking at *confusions* "
      ],
      "metadata": {
        "id": "5mxZRPH1yB3y"
      },
      "id": "5mxZRPH1yB3y"
    },
    {
      "cell_type": "code",
      "execution_count": 79,
      "id": "326f745a-92ed-46f5-9257-02b218000930",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T02:16:56.874050Z",
          "iopub.status.busy": "2023-05-04T02:16:56.872819Z",
          "iopub.status.idle": "2023-05-04T02:16:57.067166Z",
          "shell.execute_reply": "2023-05-04T02:16:57.062981Z",
          "shell.execute_reply.started": "2023-05-04T02:16:56.874020Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 447
        },
        "id": "326f745a-92ed-46f5-9257-02b218000930",
        "outputId": "a09a5a89-6b30-428c-c301-40a5c9acd359"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7fe071901660>"
            ]
          },
          "metadata": {},
          "execution_count": 79
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "a_coat, a_coat_label = test_data[6]\n",
        "plt.imshow(a_coat.squeeze())"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "776fdef4-61c4-4f3f-957b-17001f0782f3",
      "metadata": {
        "id": "776fdef4-61c4-4f3f-957b-17001f0782f3"
      },
      "source": [
        "This coat is so light, it looks closer to the average sandal in the pixel comparison"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "coat_distances = []\n",
        "\n",
        "for i in range(10):\n",
        "  coat_distances.append(mnist_distance(a_coat, mean_items[i]))\n",
        "\n",
        "print(f'\\nBaseline with pixel comparison\\nPredicted: {classes[torch.argmin(torch.stack(coat_distances))]}, Actual: {classes[a_coat_label]}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "A5j9zIaMv4vt",
        "outputId": "8fbcd22a-3438-4dfd-9c0b-d1d2b6bb2876"
      },
      "id": "A5j9zIaMv4vt",
      "execution_count": 69,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Baseline with pixel comparison\n",
            "Predicted: Sandal, Actual: Coat\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "---\n",
        "\n",
        "# Using a neural network"
      ],
      "metadata": {
        "id": "9lfJSQ9GyI_A"
      },
      "id": "9lfJSQ9GyI_A"
    },
    {
      "cell_type": "code",
      "execution_count": 70,
      "id": "3191c09f-9baa-441d-a41f-ab1e2230ed28",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T01:15:50.301802Z",
          "iopub.status.busy": "2023-05-04T01:15:50.301536Z",
          "iopub.status.idle": "2023-05-04T01:15:53.400207Z",
          "shell.execute_reply": "2023-05-04T01:15:53.398817Z",
          "shell.execute_reply.started": "2023-05-04T01:15:50.301778Z"
        },
        "id": "3191c09f-9baa-441d-a41f-ab1e2230ed28"
      },
      "outputs": [],
      "source": [
        "class NeuralNetwork(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(NeuralNetwork, self).__init__()\n",
        "        self.flatten = nn.Flatten()\n",
        "        self.linear_relu_stack = nn.Sequential(\n",
        "            nn.Linear(28*28, 512),\n",
        "            nn.ReLU(),\n",
        "            nn.Linear(512, 512),\n",
        "            nn.ReLU(),\n",
        "            nn.Linear(512, 10),\n",
        "        )\n",
        "        \n",
        "    def forward(self, x):\n",
        "        x = self.flatten(x)\n",
        "        logits = self.linear_relu_stack(x)\n",
        "        return logits\n",
        "\n",
        "model = NeuralNetwork().to(device)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 71,
      "id": "0473cdf4-6bf5-4865-a984-1fcaf47714a4",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T01:15:53.402376Z",
          "iopub.status.busy": "2023-05-04T01:15:53.401942Z",
          "iopub.status.idle": "2023-05-04T01:15:53.407708Z",
          "shell.execute_reply": "2023-05-04T01:15:53.406660Z",
          "shell.execute_reply.started": "2023-05-04T01:15:53.402348Z"
        },
        "id": "0473cdf4-6bf5-4865-a984-1fcaf47714a4"
      },
      "outputs": [],
      "source": [
        "loss_fn = nn.CrossEntropyLoss()\n",
        "optimizer = torch.optim.SGD(model.parameters(), lr=1e-3)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 72,
      "id": "821c535b-516d-4640-b4e2-fa62d6e8732c",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T01:16:01.152202Z",
          "iopub.status.busy": "2023-05-04T01:16:01.151826Z",
          "iopub.status.idle": "2023-05-04T01:16:01.159428Z",
          "shell.execute_reply": "2023-05-04T01:16:01.158405Z",
          "shell.execute_reply.started": "2023-05-04T01:16:01.152129Z"
        },
        "id": "821c535b-516d-4640-b4e2-fa62d6e8732c"
      },
      "outputs": [],
      "source": [
        "def train(dataloader, model, loss_fn, optimizer):\n",
        "    size = len(dataloader.dataset)\n",
        "    model.train()\n",
        "    for batch, (X, y) in enumerate(dataloader):\n",
        "        if device != \"cpu\":\n",
        "            X, y = X.to(device), y.to(device)\n",
        "        \n",
        "        pred = model(X)\n",
        "        loss = loss_fn(pred, y)\n",
        "        \n",
        "        optimizer.zero_grad()\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "        \n",
        "        if batch % 100 == 0:\n",
        "            loss, current = loss.item(), (batch + 1) * len(X)\n",
        "            print(f\"loss: {loss:>7f}, [{current:>5d}/{size:>5d}]\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 73,
      "id": "937766f0-ba3a-49dd-9b61-9caf83d32823",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T01:20:17.795536Z",
          "iopub.status.busy": "2023-05-04T01:20:17.793652Z",
          "iopub.status.idle": "2023-05-04T01:20:17.805341Z",
          "shell.execute_reply": "2023-05-04T01:20:17.803395Z",
          "shell.execute_reply.started": "2023-05-04T01:20:17.795522Z"
        },
        "id": "937766f0-ba3a-49dd-9b61-9caf83d32823"
      },
      "outputs": [],
      "source": [
        "def test(dataloader, model, loss_fn):\n",
        "    size = len(dataloader.dataset)\n",
        "    num_batches = len(dataloader)\n",
        "    model.eval()\n",
        "    test_loss, correct = 0, 0\n",
        "    with torch.no_grad():\n",
        "        for X, y in dataloader:\n",
        "            X, y = X.to(device), y.to(device)\n",
        "            pred = model(X)\n",
        "            test_loss += loss_fn(pred, y).item()\n",
        "            correct += (pred.argmax(1) == y).type(torch.float).sum().item()\n",
        "    test_loss /= num_batches\n",
        "    correct /= size\n",
        "    print(f\"Test error: \\n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f} \\n\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 74,
      "id": "5f17a1dc-8f45-41de-9b71-187dc8c8da76",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T01:15:50.294663Z",
          "iopub.status.busy": "2023-05-04T01:15:50.294348Z",
          "iopub.status.idle": "2023-05-04T01:15:50.300430Z",
          "shell.execute_reply": "2023-05-04T01:15:50.299541Z",
          "shell.execute_reply.started": "2023-05-04T01:15:50.294635Z"
        },
        "id": "5f17a1dc-8f45-41de-9b71-187dc8c8da76"
      },
      "outputs": [],
      "source": [
        "train_dataloader = DataLoader(train_data, batch_size=64)\n",
        "test_dataloader = DataLoader(test_data, batch_size=64)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 24,
      "id": "75c980da-1552-483f-ba44-d3fa39648336",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2023-05-04T01:20:18.866036Z",
          "iopub.status.busy": "2023-05-04T01:20:18.865696Z",
          "iopub.status.idle": "2023-05-04T01:21:12.293972Z",
          "shell.execute_reply": "2023-05-04T01:21:12.292670Z",
          "shell.execute_reply.started": "2023-05-04T01:20:18.866011Z"
        },
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "75c980da-1552-483f-ba44-d3fa39648336",
        "outputId": "03d1e177-f13f-4189-dc20-95ddcdc8c579"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1\n",
            "-------------------------------\n",
            "loss: 2.311673, [   64/60000]\n",
            "loss: 2.299227, [ 6464/60000]\n",
            "loss: 2.281137, [12864/60000]\n",
            "loss: 2.263207, [19264/60000]\n",
            "loss: 2.245568, [25664/60000]\n",
            "loss: 2.232702, [32064/60000]\n",
            "loss: 2.222782, [38464/60000]\n",
            "loss: 2.201670, [44864/60000]\n",
            "loss: 2.190649, [51264/60000]\n",
            "loss: 2.168225, [57664/60000]\n",
            "Test error: \n",
            " Accuracy: 48.2%, Avg loss: 2.157065 \n",
            "\n",
            "Epoch 2\n",
            "-------------------------------\n",
            "loss: 2.168553, [   64/60000]\n",
            "loss: 2.155025, [ 6464/60000]\n",
            "loss: 2.096412, [12864/60000]\n",
            "loss: 2.110275, [19264/60000]\n",
            "loss: 2.053044, [25664/60000]\n",
            "loss: 2.018328, [32064/60000]\n",
            "loss: 2.030789, [38464/60000]\n",
            "loss: 1.961182, [44864/60000]\n",
            "loss: 1.955887, [51264/60000]\n",
            "loss: 1.900697, [57664/60000]\n",
            "Test error: \n",
            " Accuracy: 58.7%, Avg loss: 1.886838 \n",
            "\n",
            "Epoch 3\n",
            "-------------------------------\n",
            "loss: 1.918088, [   64/60000]\n",
            "loss: 1.884685, [ 6464/60000]\n",
            "loss: 1.763732, [12864/60000]\n",
            "loss: 1.812748, [19264/60000]\n",
            "loss: 1.689504, [25664/60000]\n",
            "loss: 1.666705, [32064/60000]\n",
            "loss: 1.678045, [38464/60000]\n",
            "loss: 1.586120, [44864/60000]\n",
            "loss: 1.598803, [51264/60000]\n",
            "loss: 1.507476, [57664/60000]\n",
            "Test error: \n",
            " Accuracy: 61.1%, Avg loss: 1.520704 \n",
            "\n",
            "Epoch 4\n",
            "-------------------------------\n",
            "loss: 1.586154, [   64/60000]\n",
            "loss: 1.551346, [ 6464/60000]\n",
            "loss: 1.395454, [12864/60000]\n",
            "loss: 1.476769, [19264/60000]\n",
            "loss: 1.345945, [25664/60000]\n",
            "loss: 1.362685, [32064/60000]\n",
            "loss: 1.364179, [38464/60000]\n",
            "loss: 1.298594, [44864/60000]\n",
            "loss: 1.324254, [51264/60000]\n",
            "loss: 1.229019, [57664/60000]\n",
            "Test error: \n",
            " Accuracy: 62.8%, Avg loss: 1.259957 \n",
            "\n",
            "Epoch 5\n",
            "-------------------------------\n",
            "loss: 1.338718, [   64/60000]\n",
            "loss: 1.320556, [ 6464/60000]\n",
            "loss: 1.148245, [12864/60000]\n",
            "loss: 1.261043, [19264/60000]\n",
            "loss: 1.127651, [25664/60000]\n",
            "loss: 1.169615, [32064/60000]\n",
            "loss: 1.175436, [38464/60000]\n",
            "loss: 1.123663, [44864/60000]\n",
            "loss: 1.154510, [51264/60000]\n",
            "loss: 1.069890, [57664/60000]\n",
            "Test error: \n",
            " Accuracy: 64.4%, Avg loss: 1.099178 \n",
            "\n",
            "Done!\n"
          ]
        }
      ],
      "source": [
        "epochs = 5\n",
        "for t in range(epochs):\n",
        "    print(f\"Epoch {t+1}\\n-------------------------------\")\n",
        "    train(train_dataloader, model, loss_fn, optimizer)\n",
        "    test(test_dataloader, model, loss_fn)\n",
        "print(\"Done!\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 83,
      "id": "07487959-2160-448d-ac62-147d1e44a313",
      "metadata": {
        "collapsed": true,
        "execution": {
          "iopub.execute_input": "2023-05-04T02:25:11.729850Z",
          "iopub.status.busy": "2023-05-04T02:25:11.729466Z",
          "iopub.status.idle": "2023-05-04T02:25:11.749004Z",
          "shell.execute_reply": "2023-05-04T02:25:11.747931Z",
          "shell.execute_reply.started": "2023-05-04T02:25:11.729821Z"
        },
        "jupyter": {
          "outputs_hidden": true
        },
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "07487959-2160-448d-ac62-147d1e44a313",
        "outputId": "04595990-611f-4f0d-a484-3dafa6d9b6a4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Saved PyTorch Model State to model.pth\n"
          ]
        }
      ],
      "source": [
        "torch.save(model.state_dict(), \"model.pth\")\n",
        "print(\"Saved PyTorch Model State to model.pth\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 93,
      "id": "56ecc636-5d1a-4bdf-a6c9-6827d7e064b9",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "56ecc636-5d1a-4bdf-a6c9-6827d7e064b9",
        "outputId": "2a4a4d17-f4ab-4048-b193-d0750fd786d0"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<All keys matched successfully>"
            ]
          },
          "metadata": {},
          "execution_count": 93
        }
      ],
      "source": [
        "model = NeuralNetwork()\n",
        "model.load_state_dict(torch.load(\"model.pth\"))"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "980deafd-04fd-410c-a7cd-0bfcee0fb379",
      "metadata": {
        "id": "980deafd-04fd-410c-a7cd-0bfcee0fb379"
      },
      "source": [
        "## Next experiments\n",
        "\n",
        "- Using the fast.ai Learner\n",
        "- Batching the baseline calculations"
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3 (ipykernel)",
      "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": "3.9.13"
    },
    "papermill": {
      "default_parameters": {},
      "duration": 24.424144,
      "end_time": "2023-01-20T22:38:47.310506",
      "environment_variables": {},
      "exception": null,
      "input_path": "__notebook__.ipynb",
      "output_path": "__notebook__.ipynb",
      "parameters": {},
      "start_time": "2023-01-20T22:38:22.886362",
      "version": "2.3.4"
    },
    "colab": {
      "provenance": []
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "nbformat": 4,
  "nbformat_minor": 5
}