cap05/C05_E1_Shallow-AutoEncoder.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "81541504-694b-42af-88a5-2764559d1745",
   "metadata": {},
   "source": [
    "<div style=\"width: 100%; clear: both;\">\n",
    "    <div style=\"float: left; width: 50%;\">\n",
    "        <img src=\"http://www.uoc.edu/portal/_resources/common/imatges/marca_UOC/UOC_Masterbrand.jpg\", align=\"left\">\n",
    "    </div>\n",
    "    <div style=\"float: right; width: 50%;\">\n",
    "        <p style=\"margin: 0; padding-top: 22px; text-align:right;\">M2.875 - Deep Learning</p>\n",
    "        <p style=\"margin: 0; text-align:right;\">Estudios de Informática, Multimedia y Telecomunicación</p>\n",
    "        <p style=\"margin: 0; text-align:right; padding-button: 100px;\">Universitat Oberta de Catalunya (UOC)</p>\n",
    "    </div>\n",
    "</div>\n",
    "<div style=\"width:100%;\">&nbsp;</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ddb83b2f-8daf-4dad-ab23-594dc4e17ec3",
   "metadata": {},
   "source": [
    "# Ejemplo de funcionamiento de Autoencoder\n",
    "\n",
    "En este ejemplo veremos un caso de aplicación para un autoencoder. En concreto, nos centraremos en la compresión y reconstrucción de imágenes empleando *autoencoders*.\n",
    "\n",
    "<u>Nota</u>: Este ejemplo está basado en https://blog.keras.io/building-autoencoders-in-keras.html"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a1bfafd4-c1d3-4f1a-9a47-a2f94ed0fbbb",
   "metadata": {},
   "source": [
    "Deshabilitamos la aparición de warnings.\n",
    "\n",
    "<u>Nota</u>: No se recomienda este paso cuando se está desarrollando el código."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "5dbf7aae-5649-4c04-b69f-dc804439f80f",
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3fe0fe6a-892e-4c89-8a94-9fb01cbdc647",
   "metadata": {},
   "source": [
    "## Carga de las librerías\n",
    "\n",
    "Carga de las librerías para ejecutar el ejemplo."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "b5cff874-9ecd-44c9-87fe-49a444f7a14d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TensorFlow version: 2.7.0\n",
      "Keras version     : 2.7.0\n",
      "Numpy version     : 1.21.5\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "import numpy as np\n",
    "\n",
    "print(\"TensorFlow version: {}\".format(tf.__version__))\n",
    "print(\"Keras version     : {}\".format(keras.__version__))\n",
    "print(\"Numpy version     : {}\".format(np.__version__))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "1cc36661",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.utils import to_categorical\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Conv2D, Flatten\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "36927297-2d2e-47eb-986a-f488d774af21",
   "metadata": {},
   "source": [
    "## 1. Carga del conjunto de datos\n",
    "\n",
    "En primer lugar, cargamos el conjunto de datos, que en este caso será el dataset de dígitos [MNIST](https://en.wikipedia.org/wiki/MNIST_database). \n",
    "\n",
    "Realizamos esta carga, directamente, a partir de la librería [Keras](https://keras.io/)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "06d1fb57",
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import mnist\n",
    "import numpy as np\n",
    "\n",
    "(x_train, _), (x_test, _) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "1648f3c1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 784)\n",
      "(10000, 784)\n"
     ]
    }
   ],
   "source": [
    "x_train = x_train.astype('float32') / 255.\n",
    "x_test = x_test.astype('float32') / 255.\n",
    "x_train = x_train.reshape((len(x_train), np.prod(x_train.shape[1:])))\n",
    "x_test = x_test.reshape((len(x_test), np.prod(x_test.shape[1:])))\n",
    "print(x_train.shape)\n",
    "print(x_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "98459ae4-77b9-422d-8442-dae5a4f37412",
   "metadata": {},
   "source": [
    "## 2. Creación del *autoencoder*\n",
    "\n",
    "A continuación, creamos el *autoencoder* en su forma más senzilla:\n",
    "- A partir de las imágenes de entrada de de 28x28 píxeles en escala de grises, se crea una capa con 784 valores de entrada\n",
    "- Creamos un **codificador** (*encoder*) con una representación de 32 neuronas\n",
    "- Creamos un **decodificador** (*decoder*) que genere una salida en la misma resolución que las imágenes de entrada, es decir, 784 neuronas de salida para generar una imagen de 28x28 píxeles en escala de grises."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "ce4e772e",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-03-06 16:30:33.886782: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Input, Dense\n",
    "from keras.models import Model\n",
    "\n",
    "# Tamaño de la representación interna del autoencoder\n",
    "layer1dim = 64\n",
    "layer2dim = 32\n",
    "layer3dim = 64\n",
    "\n",
    "# capa de entrada\n",
    "input_img = Input(shape=(784,))\n",
    "# representación interna (encoded)\n",
    "h1 = Dense(layer1dim, activation='relu')(input_img)\n",
    "h2 = Dense(layer2dim, activation='relu')(h1)\n",
    "h3 = Dense(layer3dim, activation='relu')(h2)\n",
    "\n",
    "# salida del autoencoder, es decir, la imagen reconstruida (decoded)\n",
    "decoded = Dense(784, activation='sigmoid')(h3)\n",
    "\n",
    "# el modelo completo de autoencoder, que apila la entrada y la salida\n",
    "autoencoder = Model(input_img, decoded)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3492cd0a-946e-4234-9959-57c05a590d49",
   "metadata": {},
   "source": [
    "Para finalidades didácticas, definimos el modelo `encoder` que genera la codificación interna del autoencoder, sin el proceso de decodificación."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "fd1eb2d7",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_1 (InputLayer)        [(None, 784)]             0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 64)                50240     \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 32)                2080      \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 64)                2112      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 54,432\n",
      "Trainable params: 54,432\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# este modelo crea la representación interna\n",
    "encoder = Model(input_img, h3)\n",
    "\n",
    "encoder.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "39fd7686-98e8-4c4a-be81-00ad620d3c1a",
   "metadata": {},
   "source": [
    "De forma similar, definimos el modelo `decoder` que a partir de la representación interna (32 valores) reconstruye la imagen, es decir, aplica el proceso de decodificación."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "62dd712f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_2 (InputLayer)        [(None, 64)]              0         \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 784)               50960     \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 50,960\n",
      "Trainable params: 50,960\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# capa de entrada de la representación interna\n",
    "encoded_input = Input(shape=(layer1dim,))\n",
    "\n",
    "# obtener la última capa del autoencoder definido previamente\n",
    "decoder_layer = autoencoder.layers[-1]\n",
    "\n",
    "# creación del modelo\n",
    "decoder = Model(encoded_input, decoder_layer(encoded_input))\n",
    "\n",
    "decoder.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "04e79b95-e6d2-4682-9346-e96ebd9a7e41",
   "metadata": {},
   "source": [
    "Finalmente, definimos el optimizador y la función de pérdida empleada para el entrenamiento del *autoencoder*."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "fc3b494c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_1 (InputLayer)        [(None, 784)]             0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 64)                50240     \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 32)                2080      \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 64)                2112      \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 784)               50960     \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 105,392\n",
      "Trainable params: 105,392\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "autoencoder.compile(loss=\"binary_crossentropy\",optimizer=keras.optimizers.SGD(lr=1.5))\n",
    "autoencoder.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "64ffd802-72ba-4657-97a8-c075d9ac5c07",
   "metadata": {},
   "source": [
    "## 3. Entrenamiento del *autoencoder*\n",
    "\n",
    "En el siguiente fragmento de código se entrena el *autoencoder* creado anteriormente.\n",
    "\n",
    "Los hiperparámetros para ajustar el entrenamiento, además del optimizador y función de pérdida, son:\n",
    "- número de épocas del entrenamiento\n",
    "- tamaño del lote"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "56bfcd72",
   "metadata": {},
   "outputs": [],
   "source": [
    "n_epochs = 50\n",
    "n_batch_size = 256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "0ba7484d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.3031 - val_loss: 0.2625\n",
      "Epoch 2/50\n",
      "235/235 [==============================] - 1s 3ms/step - loss: 0.2402 - val_loss: 0.2166\n",
      "Epoch 3/50\n",
      "235/235 [==============================] - 1s 3ms/step - loss: 0.2043 - val_loss: 0.1939\n",
      "Epoch 4/50\n",
      "235/235 [==============================] - 1s 3ms/step - loss: 0.1866 - val_loss: 0.1798\n",
      "Epoch 5/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1732 - val_loss: 0.1642\n",
      "Epoch 6/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1628 - val_loss: 0.1576\n",
      "Epoch 7/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1563 - val_loss: 0.1528\n",
      "Epoch 8/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1517 - val_loss: 0.1466\n",
      "Epoch 9/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1480 - val_loss: 0.1494\n",
      "Epoch 10/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1438 - val_loss: 0.1413\n",
      "Epoch 11/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1395 - val_loss: 0.1386\n",
      "Epoch 12/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1363 - val_loss: 0.1330\n",
      "Epoch 13/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1338 - val_loss: 0.1343\n",
      "Epoch 14/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1316 - val_loss: 0.1274\n",
      "Epoch 15/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1295 - val_loss: 0.1297\n",
      "Epoch 16/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1279 - val_loss: 0.1236\n",
      "Epoch 17/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1263 - val_loss: 0.1247\n",
      "Epoch 18/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1249 - val_loss: 0.1220\n",
      "Epoch 19/50\n",
      "235/235 [==============================] - 1s 4ms/step - loss: 0.1233 - val_loss: 0.1215\n",
      "Epoch 20/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1221 - val_loss: 0.1201\n",
      "Epoch 21/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1208 - val_loss: 0.1172\n",
      "Epoch 22/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1196 - val_loss: 0.1201\n",
      "Epoch 23/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1188 - val_loss: 0.1173\n",
      "Epoch 24/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1177 - val_loss: 0.1168\n",
      "Epoch 25/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1169 - val_loss: 0.1141\n",
      "Epoch 26/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1160 - val_loss: 0.1142\n",
      "Epoch 27/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1154 - val_loss: 0.1145\n",
      "Epoch 28/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1146 - val_loss: 0.1150\n",
      "Epoch 29/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1140 - val_loss: 0.1117\n",
      "Epoch 30/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1133 - val_loss: 0.1136\n",
      "Epoch 31/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1124 - val_loss: 0.1107\n",
      "Epoch 32/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1118 - val_loss: 0.1093\n",
      "Epoch 33/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1112 - val_loss: 0.1105\n",
      "Epoch 34/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1107 - val_loss: 0.1093\n",
      "Epoch 35/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1099 - val_loss: 0.1065\n",
      "Epoch 36/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1093 - val_loss: 0.1088\n",
      "Epoch 37/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1089 - val_loss: 0.1065\n",
      "Epoch 38/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1084 - val_loss: 0.1081\n",
      "Epoch 39/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1080 - val_loss: 0.1044\n",
      "Epoch 40/50\n",
      "235/235 [==============================] - 1s 5ms/step - loss: 0.1076 - val_loss: 0.1069\n",
      "Epoch 41/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1072 - val_loss: 0.1046\n",
      "Epoch 42/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1067 - val_loss: 0.1066\n",
      "Epoch 43/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1066 - val_loss: 0.1055\n",
      "Epoch 44/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1060 - val_loss: 0.1065\n",
      "Epoch 45/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1057 - val_loss: 0.1036\n",
      "Epoch 46/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1052 - val_loss: 0.1023\n",
      "Epoch 47/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1050 - val_loss: 0.1045\n",
      "Epoch 48/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1044 - val_loss: 0.1020\n",
      "Epoch 49/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1041 - val_loss: 0.1004\n",
      "Epoch 50/50\n",
      "235/235 [==============================] - 1s 6ms/step - loss: 0.1038 - val_loss: 0.1029\n"
     ]
    }
   ],
   "source": [
    "mfit = autoencoder.fit(x_train, x_train,\n",
    "                epochs=n_epochs,\n",
    "                batch_size=n_batch_size,\n",
    "                shuffle=True,\n",
    "                validation_data=(x_test, x_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "adc1d5ba",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_3\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_3 (InputLayer)        [(None, 28, 28)]          0         \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 784)               0         \n",
      "                                                                 \n",
      " dense_4 (Dense)             (None, 392)               307720    \n",
      "                                                                 \n",
      " dense_5 (Dense)             (None, 196)               77028     \n",
      "                                                                 \n",
      " dense_6 (Dense)             (None, 392)               77224     \n",
      "                                                                 \n",
      " dense_7 (Dense)             (None, 784)               308112    \n",
      "                                                                 \n",
      " reshape (Reshape)           (None, 28, 28)            0         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 770,084\n",
      "Trainable params: 770,084\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#Stacked Autoencoder with functional model\n",
    "#encoder\n",
    "inputs = keras.Input(shape=(28,28))\n",
    "lr_flatten = keras.layers.Flatten()(inputs)\n",
    "lr1 = keras.layers.Dense(392, activation=\"selu\")(lr_flatten)\n",
    "lr2 = keras.layers.Dense(196, activation=\"selu\")(lr1)\n",
    "#decoder\n",
    "lr3 =  keras.layers.Dense(392, activation=\"selu\")(lr2)\n",
    "lr4 =  keras.layers.Dense(28 * 28, activation=\"sigmoid\")(lr3)\n",
    "outputs = keras.layers.Reshape([28, 28])(lr4)\n",
    "stacked_ae = keras.models.Model(inputs,outputs)\n",
    "stacked_ae.compile(loss=\"binary_crossentropy\", optimizer=keras.optimizers.SGD(lr=1.5))\n",
    "stacked_ae.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "de28c971",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot del training loss\n",
    "plt.style.use(\"ggplot\")\n",
    "\n",
    "plt.plot(np.arange(0, n_epochs), mfit.history[\"loss\"], label=\"train\")\n",
    "plt.plot(np.arange(0, n_epochs), mfit.history[\"val_loss\"], label=\"val\")\n",
    "plt.title(\"Loss\")\n",
    "plt.xlabel(\"Epoch #\")\n",
    "plt.ylabel(\"Loss\")\n",
    "plt.legend(loc=\"upper right\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ab78130c-add0-4ef2-83f9-bbc5f58fc675",
   "metadata": {},
   "source": [
    "## 4. Visualización de los resultados\n",
    "\n",
    "Finalmente, presentaremos los resultados de aplicar el *autoencoder* en algunos ejemplos del conjunto de datos de test."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "79d8f608",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# codificamos y decodificamos las imágenes de test\n",
    "encoded_imgs = encoder.predict(x_test)\n",
    "decoded_imgs = decoder.predict(encoded_imgs)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "618d14c8-3c7a-4549-b6bc-56c595fd9e6d",
   "metadata": {},
   "source": [
    "Notese que este paso se puede realizar en un solo paso, empleando:\n",
    "\n",
    "> decoded_imgs = autoencoder.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "e3f9c527",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2000x400 with 20 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "n = 10  # cantidad de ejemplos que mostrar\n",
    "plt.figure(figsize=(20, 4))\n",
    "\n",
    "for i in range(n):\n",
    "    # mostrar la imagen original\n",
    "    ax = plt.subplot(2, n, i + 1)\n",
    "    plt.imshow(x_test[i].reshape(28, 28))\n",
    "    plt.gray()\n",
    "    ax.get_xaxis().set_visible(False)\n",
    "    ax.get_yaxis().set_visible(False)\n",
    "\n",
    "    # mostrar la reconstrucción\n",
    "    ax = plt.subplot(2, n, i + 1 + n)\n",
    "    plt.imshow(decoded_imgs[i].reshape(28, 28))\n",
    "    plt.gray()\n",
    "    ax.get_xaxis().set_visible(False)\n",
    "    ax.get_yaxis().set_visible(False)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e572ba02",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}