machinelearning

Machine learning code
git clone git://git.laack.co/machinelearning.git
Log | Files | Refs

MNISTAutoencoder.ipynb (13518B)

      1 {
      2  "cells": [
      3   {
      4    "cell_type": "markdown",
      5    "metadata": {},
      6    "source": [
      7     "This is an autoencoder trained to compress mnist hand written digits. "
      8    ]
      9   },
     10   {
     11    "cell_type": "code",
     12    "execution_count": 85,
     13    "metadata": {},
     14    "outputs": [],
     15    "source": [
     16     "from sklearn.datasets import fetch_openml\n",
     17     "import matplotlib.pyplot as plt\n",
     18     "import numpy as np\n",
     19     "from sklearn.preprocessing import MinMaxScaler\n",
     20     "\n",
     21     "# Fetch the MNIST dataset\n",
     22     "mnist = fetch_openml(\"mnist_784\", as_frame=False)\n",
     23     "\n",
     24     "minMax = MinMaxScaler()\n",
     25     "\n",
     26     "\n",
     27     "# Extract data and labels\n",
     28     "X, y = mnist.data, mnist.target\n",
     29     "\n",
     30     "X = minMax.fit_transform(X)\n",
     31     "\n",
     32     "# Reshape images to 28x28\n",
     33     "X = X.reshape(-1, 28, 28)\n",
     34     "\n",
     35     "# Split the data into training and test sets\n",
     36     "X_train, X_test, y_train, y_test = X[:60000], X[60000:], y[:60000], y[60000:]\n",
     37     "\n",
     38     "# Function to plot an image\n",
     39     "def plot(image_data):\n",
     40     "    plt.imshow(image_data, cmap='binary')\n",
     41     "    plt.axis('off')\n",
     42     "    plt.show()\n"
     43    ]
     44   },
     45   {
     46    "cell_type": "code",
     47    "execution_count": 86,
     48    "metadata": {},
     49    "outputs": [],
     50    "source": [
     51     "import keras\n",
     52     "import tensorflow as tf \n",
     53     "\n",
     54     "\n",
     55     "input_layer = keras.layers.Input(shape=(28,28,1))\n",
     56     "flatten_layer = keras.layers.Flatten()\n",
     57     "first_layer = keras.layers.Dense(784, activation='relu')\n",
     58     "second_layer = keras.layers.Dense(256, activation='relu')\n",
     59     "third_layer = keras.layers.Dense(128, activation='relu')\n",
     60     "fourth_layer = keras.layers.Dense(256, activation='relu')\n",
     61     "fifth_layer = keras.layers.Dense(784, activation='sigmoid')\n",
     62     "unflatten_layer = keras.layers.Reshape(target_shape=(28,28,1))\n"
     63    ]
     64   },
     65   {
     66    "cell_type": "code",
     67    "execution_count": 87,
     68    "metadata": {},
     69    "outputs": [],
     70    "source": [
     71     "autoencoder = keras.Sequential(layers=[\n",
     72     "    input_layer,\n",
     73     "    flatten_layer,\n",
     74     "    first_layer,\n",
     75     "    second_layer,\n",
     76     "    third_layer,\n",
     77     "    fourth_layer,\n",
     78     "    fifth_layer,\n",
     79     "    unflatten_layer\n",
     80     "])"
     81    ]
     82   },
     83   {
     84    "cell_type": "code",
     85    "execution_count": 88,
     86    "metadata": {},
     87    "outputs": [],
     88    "source": [
     89     "autoencoder.compile(loss=keras.losses.MeanSquaredError, optimizer='adam')"
     90    ]
     91   },
     92   {
     93    "cell_type": "code",
     94    "execution_count": 89,
     95    "metadata": {},
     96    "outputs": [
     97     {
     98      "name": "stdout",
     99      "output_type": "stream",
    100      "text": [
    101       "Epoch 1/10\n",
    102       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 7ms/step - loss: 0.0315\n",
    103       "Epoch 2/10\n",
    104       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 6ms/step - loss: 0.0074\n",
    105       "Epoch 3/10\n",
    106       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 7ms/step - loss: 0.0057\n",
    107       "Epoch 4/10\n",
    108       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 6ms/step - loss: 0.0049\n",
    109       "Epoch 5/10\n",
    110       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 7ms/step - loss: 0.0044\n",
    111       "Epoch 6/10\n",
    112       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 7ms/step - loss: 0.0041\n",
    113       "Epoch 7/10\n",
    114       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 6ms/step - loss: 0.0038\n",
    115       "Epoch 8/10\n",
    116       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 6ms/step - loss: 0.0036\n",
    117       "Epoch 9/10\n",
    118       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 6ms/step - loss: 0.0034\n",
    119       "Epoch 10/10\n",
    120       "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 6ms/step - loss: 0.0033\n"
    121      ]
    122     },
    123     {
    124      "data": {
    125       "text/plain": [
    126        "<keras.src.callbacks.history.History at 0x7f225c1903d0>"
    127       ]
    128      },
    129      "execution_count": 89,
    130      "metadata": {},
    131      "output_type": "execute_result"
    132     }
    133    ],
    134    "source": [
    135     "autoencoder.fit(X_train, X_train, epochs=10)"
    136    ]
    137   },
    138   {
    139    "cell_type": "code",
    140    "execution_count": 90,
    141    "metadata": {},
    142    "outputs": [],
    143    "source": [
    144     "encoder = keras.Sequential(layers=[\n",
    145     "    input_layer,\n",
    146     "    flatten_layer,\n",
    147     "    first_layer,\n",
    148     "    second_layer,\n",
    149     "    third_layer,\n",
    150     "])\n",
    151     "\n",
    152     "decoder = keras.Sequential(layers=[\n",
    153     "    third_layer,\n",
    154     "    fourth_layer,\n",
    155     "    fifth_layer,\n",
    156     "    unflatten_layer\n",
    157     "])"
    158    ]
    159   },
    160   {
    161    "cell_type": "code",
    162    "execution_count": 92,
    163    "metadata": {},
    164    "outputs": [
    165     {
    166      "name": "stdout",
    167      "output_type": "stream",
    168      "text": [
    169       "\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 14ms/step\n",
    170       "Before:\n"
    171      ]
    172     },
    173     {
    174      "data": {
    175       "image/png": "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",
    176       "text/plain": [
    177        "<Figure size 640x480 with 1 Axes>"
    178       ]
    179      },
    180      "metadata": {},
    181      "output_type": "display_data"
    182     },
    183     {
    184      "name": "stdout",
    185      "output_type": "stream",
    186      "text": [
    187       "After:\n"
    188      ]
    189     },
    190     {
    191      "data": {
    192       "image/png": "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",
    193       "text/plain": [
    194        "<Figure size 640x480 with 1 Axes>"
    195       ]
    196      },
    197      "metadata": {},
    198      "output_type": "display_data"
    199     }
    200    ],
    201    "source": [
    202     "prediction = autoencoder.predict(np.array([X_test[0]]))\n",
    203     "print('Before:')\n",
    204     "plot(prediction[0])\n",
    205     "print('After:')\n",
    206     "plot(X_test[0])"
    207    ]
    208   },
    209   {
    210    "cell_type": "markdown",
    211    "metadata": {},
    212    "source": [
    213     "These look virtualy identical and the second one was compressed into 128 dimensions instead of the original 784."
    214    ]
    215   }
    216  ],
    217  "metadata": {
    218   "kernelspec": {
    219    "display_name": ".venv",
    220    "language": "python",
    221    "name": "python3"
    222   },
    223   "language_info": {
    224    "codemirror_mode": {
    225     "name": "ipython",
    226     "version": 3
    227    },
    228    "file_extension": ".py",
    229    "mimetype": "text/x-python",
    230    "name": "python",
    231    "nbconvert_exporter": "python",
    232    "pygments_lexer": "ipython3",
    233    "version": "3.11.2"
    234   }
    235  },
    236  "nbformat": 4,
    237  "nbformat_minor": 2
    238 }