LEBLANC_TD4.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "LEBLANC_TD4.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MmVBfrAJqQ_e"
      },
      "source": [
        "# Part 1"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z0Gt32f53Tb6"
      },
      "source": [
        "## Exercise 1"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1ao8i30c1mFw",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7e385a73-a67f-4c90-c79e-09639cda8ada"
      },
      "source": [
        "from keras.datasets import imdb\n",
        "\n",
        "(train_data, train_labels), (test_data, test_labels) = imdb.load_data(num_words=10000)\n",
        "\n",
        "print(\"The train_data size is\", len(train_data), \"and the test_data size is\", len(test_data))\n",
        "print(\"The train_data shape is\", train_data.shape, \"and the test_data shape is\", test_data.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<string>:6: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/datasets/imdb.py:159: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  x_train, y_train = np.array(xs[:idx]), np.array(labels[:idx])\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "The train_data size is 25000 and the test_data size is 25000\n",
            "The train_data shape is (25000,) and the test_data shape is (25000,)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/datasets/imdb.py:160: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  x_test, y_test = np.array(xs[idx:]), np.array(labels[idx:])\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "M_wjLsf56Beu"
      },
      "source": [
        "## Exercise 2"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iZNh4DpP6DIk",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3b185241-5835-4059-8695-7462c9ead412"
      },
      "source": [
        "import numpy as np\n",
        "\n",
        "def vectorize_sequences(sequences, dimension = 10000) :\n",
        "  # Create an all-zero matrix of shape (len(sequences), dimension)\n",
        "  results = np.zeros ((len(sequences), dimension))\n",
        "  for i , sequence in enumerate (sequences):\n",
        "    results [i, sequence] = 1. # set specific indices of results [i] to 1s\n",
        "  return results\n",
        "\n",
        "# Our vectorized training data\n",
        "x_train = vectorize_sequences(train_data)\n",
        "# Our vectorized test data\n",
        "x_test = vectorize_sequences(test_data)\n",
        "# Our vectorized labels\n",
        "y_train = np.asarray (train_labels).astype('float32')\n",
        "y_test = np.asarray(test_labels).astype('float32')\n",
        "\n",
        "print(\"The x_train size is\", len(x_train), \"and the x_test size is\", len(x_test))\n",
        "print(\"The x_train shape is\", x_train.shape, \"and the x_test shape is\", x_test.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "The x_train size is 25000 and the x_test size is 25000\n",
            "The x_train shape is (25000, 10000) and the x_test shape is (25000, 10000)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kAxSm38jgQ8F"
      },
      "source": [
        "## Exercise 3"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TVhxws2GgTvd"
      },
      "source": [
        "from tensorflow import keras\n",
        "\n",
        "original_model = keras.Sequential([\n",
        "    keras.layers.Dense(16, activation='relu', input_shape=(10000,)),\n",
        "    keras.layers.Dense(16, activation='relu'),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "# For the network compilation\n",
        "original_model.compile(optimizer='rmsprop',\n",
        "                       loss= 'binary_crossentropy',\n",
        "                       metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qJq9KFLrg2zJ"
      },
      "source": [
        "The binary crossentropy loss is a sigmoid activation plus a crossentropy loss. Unlike softmax loss, it is independent for each vector component (class), meaning that the loss computed for every CNN output vector component is not affected by other component values."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tllU9cWqhDZB"
      },
      "source": [
        "## Exercise 4"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OpF7Iw6PhGtg"
      },
      "source": [
        "smaller_model = keras.Sequential([\n",
        "    keras.layers.Dense(4, activation='relu', input_shape=(10000,)),\n",
        "    keras.layers.Dense(4, activation='relu'),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "smaller_model.compile(optimizer='rmsprop',\n",
        "                       loss= 'binary_crossentropy',\n",
        "                       metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k2_edriphtrL"
      },
      "source": [
        "## Exercise 5"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XzoPcR5hhvd5",
        "outputId": "b313e494-ae96-4b73-8488-e456844e9578"
      },
      "source": [
        "original_hist = original_model.fit(x_train, y_train,\n",
        "                                   epochs=20,\n",
        "                                   batch_size=512,\n",
        "                                   validation_data=(x_test, y_test))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "49/49 [==============================] - 4s 55ms/step - loss: 0.5400 - acc: 0.7445 - val_loss: 0.3251 - val_acc: 0.8818\n",
            "Epoch 2/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.2635 - acc: 0.9084 - val_loss: 0.2825 - val_acc: 0.8905\n",
            "Epoch 3/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.1956 - acc: 0.9340 - val_loss: 0.2812 - val_acc: 0.8891\n",
            "Epoch 4/20\n",
            "49/49 [==============================] - 1s 31ms/step - loss: 0.1584 - acc: 0.9480 - val_loss: 0.3322 - val_acc: 0.8696\n",
            "Epoch 5/20\n",
            "49/49 [==============================] - 2s 32ms/step - loss: 0.1351 - acc: 0.9543 - val_loss: 0.3169 - val_acc: 0.8788\n",
            "Epoch 6/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.1200 - acc: 0.9595 - val_loss: 0.3628 - val_acc: 0.8674\n",
            "Epoch 7/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.1144 - acc: 0.9631 - val_loss: 0.3576 - val_acc: 0.8728\n",
            "Epoch 8/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.0962 - acc: 0.9680 - val_loss: 0.3863 - val_acc: 0.8678\n",
            "Epoch 9/20\n",
            "49/49 [==============================] - 1s 31ms/step - loss: 0.0808 - acc: 0.9736 - val_loss: 0.4448 - val_acc: 0.8582\n",
            "Epoch 10/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.0732 - acc: 0.9776 - val_loss: 0.4556 - val_acc: 0.8598\n",
            "Epoch 11/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.0624 - acc: 0.9822 - val_loss: 0.4898 - val_acc: 0.8573\n",
            "Epoch 12/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.0565 - acc: 0.9841 - val_loss: 0.5021 - val_acc: 0.8592\n",
            "Epoch 13/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.0479 - acc: 0.9859 - val_loss: 0.5356 - val_acc: 0.8552\n",
            "Epoch 14/20\n",
            "49/49 [==============================] - 1s 29ms/step - loss: 0.0401 - acc: 0.9895 - val_loss: 0.5626 - val_acc: 0.8570\n",
            "Epoch 15/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.0359 - acc: 0.9908 - val_loss: 0.6213 - val_acc: 0.8496\n",
            "Epoch 16/20\n",
            "49/49 [==============================] - 1s 29ms/step - loss: 0.0294 - acc: 0.9927 - val_loss: 0.6725 - val_acc: 0.8471\n",
            "Epoch 17/20\n",
            "49/49 [==============================] - 1s 31ms/step - loss: 0.0317 - acc: 0.9912 - val_loss: 0.6804 - val_acc: 0.8502\n",
            "Epoch 18/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.0249 - acc: 0.9939 - val_loss: 0.7045 - val_acc: 0.8523\n",
            "Epoch 19/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.0186 - acc: 0.9965 - val_loss: 0.7458 - val_acc: 0.8499\n",
            "Epoch 20/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.0129 - acc: 0.9979 - val_loss: 0.7884 - val_acc: 0.8488\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Td5H_yJYiFHU"
      },
      "source": [
        "## Exercise 6"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "x89nsN3riEo8",
        "outputId": "9170e0e3-2a75-4f4b-d74f-f87608b00f4f"
      },
      "source": [
        "smaller_hist = smaller_model.fit(x_train, y_train,\n",
        "                                   epochs=20,\n",
        "                                   batch_size=512,\n",
        "                                   validation_data=(x_test, y_test))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "49/49 [==============================] - 2s 36ms/step - loss: 0.5985 - acc: 0.7397 - val_loss: 0.4353 - val_acc: 0.8575\n",
            "Epoch 2/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.3731 - acc: 0.8861 - val_loss: 0.3439 - val_acc: 0.8765\n",
            "Epoch 3/20\n",
            "49/49 [==============================] - 1s 27ms/step - loss: 0.2729 - acc: 0.9082 - val_loss: 0.2966 - val_acc: 0.8870\n",
            "Epoch 4/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.2219 - acc: 0.9264 - val_loss: 0.2813 - val_acc: 0.8900\n",
            "Epoch 5/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1876 - acc: 0.9384 - val_loss: 0.2807 - val_acc: 0.8880\n",
            "Epoch 6/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1634 - acc: 0.9482 - val_loss: 0.2850 - val_acc: 0.8858\n",
            "Epoch 7/20\n",
            "49/49 [==============================] - 1s 25ms/step - loss: 0.1416 - acc: 0.9558 - val_loss: 0.2941 - val_acc: 0.8824\n",
            "Epoch 8/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1266 - acc: 0.9593 - val_loss: 0.3049 - val_acc: 0.8811\n",
            "Epoch 9/20\n",
            "49/49 [==============================] - 1s 25ms/step - loss: 0.1172 - acc: 0.9627 - val_loss: 0.3182 - val_acc: 0.8783\n",
            "Epoch 10/20\n",
            "49/49 [==============================] - 1s 24ms/step - loss: 0.1055 - acc: 0.9671 - val_loss: 0.3330 - val_acc: 0.8764\n",
            "Epoch 11/20\n",
            "49/49 [==============================] - 1s 25ms/step - loss: 0.0963 - acc: 0.9718 - val_loss: 0.3475 - val_acc: 0.8739\n",
            "Epoch 12/20\n",
            "49/49 [==============================] - 1s 25ms/step - loss: 0.0871 - acc: 0.9749 - val_loss: 0.3625 - val_acc: 0.8728\n",
            "Epoch 13/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.0829 - acc: 0.9767 - val_loss: 0.3800 - val_acc: 0.8697\n",
            "Epoch 14/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.0747 - acc: 0.9797 - val_loss: 0.3996 - val_acc: 0.8689\n",
            "Epoch 15/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.0655 - acc: 0.9819 - val_loss: 0.4178 - val_acc: 0.8676\n",
            "Epoch 16/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.0605 - acc: 0.9841 - val_loss: 0.4425 - val_acc: 0.8630\n",
            "Epoch 17/20\n",
            "49/49 [==============================] - 1s 25ms/step - loss: 0.0553 - acc: 0.9865 - val_loss: 0.4687 - val_acc: 0.8607\n",
            "Epoch 18/20\n",
            "49/49 [==============================] - 1s 25ms/step - loss: 0.0484 - acc: 0.9889 - val_loss: 0.4803 - val_acc: 0.8628\n",
            "Epoch 19/20\n",
            "49/49 [==============================] - 1s 25ms/step - loss: 0.0454 - acc: 0.9894 - val_loss: 0.5039 - val_acc: 0.8605\n",
            "Epoch 20/20\n",
            "49/49 [==============================] - 1s 25ms/step - loss: 0.0396 - acc: 0.9909 - val_loss: 0.5409 - val_acc: 0.8555\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ku3lfk4IiQ4Q"
      },
      "source": [
        "## Exercise 7"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C6K-mRC-iSeK"
      },
      "source": [
        "original_val_loss = original_hist.history['val_loss']\n",
        "smaller_model_val_loss = smaller_hist.history['val_loss']"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1UusAZCPikYr"
      },
      "source": [
        "## Exercise 8"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "o5pGoAguil_a",
        "outputId": "fba93550-3603-4e2c-9092-0f8a4e26c2c7"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "epochs = range(1, 21)\n",
        "\n",
        "plt.plot(epochs, original_val_loss, 'b+', label='Original model')\n",
        "plt.plot(epochs, smaller_model_val_loss, 'bo', label='Smaller model')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Validation loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xAC0s_wajFbC"
      },
      "source": [
        "The original model starts overfitting after 3 epochs and the smaller model starts overfitting after 5 models (and its performance degrades more slowly once it starts overfitting)."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tiwtQFfcjiiX"
      },
      "source": [
        "## Exercise 9"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jjObCY42jj2r"
      },
      "source": [
        "bigger_model = keras.Sequential([\n",
        "    keras.layers.Dense(512, activation='relu', input_shape=(10000,)),\n",
        "    keras.layers.Dense(512, activation='relu'),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "bigger_model.compile(optimizer='rmsprop',\n",
        "                       loss= 'binary_crossentropy',\n",
        "                       metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "I_rh97fujyhq"
      },
      "source": [
        "## Exercise 10"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gm1VyNcPj4ns",
        "outputId": "ee2d5ed8-70e2-4434-e1c0-f3d803e5eda5"
      },
      "source": [
        "bigger_hist = bigger_model.fit(x_train, y_train,\n",
        "                                     epochs=20,\n",
        "                                     batch_size=512,\n",
        "                                     validation_data=(x_test, y_test))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "49/49 [==============================] - 17s 321ms/step - loss: 0.6346 - acc: 0.7287 - val_loss: 0.3869 - val_acc: 0.8190\n",
            "Epoch 2/20\n",
            "49/49 [==============================] - 14s 294ms/step - loss: 0.2286 - acc: 0.9108 - val_loss: 0.3956 - val_acc: 0.8414\n",
            "Epoch 3/20\n",
            "49/49 [==============================] - 14s 293ms/step - loss: 0.1287 - acc: 0.9506 - val_loss: 0.3483 - val_acc: 0.8778\n",
            "Epoch 4/20\n",
            "49/49 [==============================] - 15s 301ms/step - loss: 0.0522 - acc: 0.9813 - val_loss: 0.3997 - val_acc: 0.8837\n",
            "Epoch 5/20\n",
            "49/49 [==============================] - 15s 298ms/step - loss: 0.0277 - acc: 0.9940 - val_loss: 0.4930 - val_acc: 0.8828\n",
            "Epoch 6/20\n",
            "49/49 [==============================] - 14s 294ms/step - loss: 0.0023 - acc: 0.9998 - val_loss: 0.6999 - val_acc: 0.8816\n",
            "Epoch 7/20\n",
            "49/49 [==============================] - 14s 297ms/step - loss: 0.0105 - acc: 0.9989 - val_loss: 0.7612 - val_acc: 0.8530\n",
            "Epoch 8/20\n",
            "49/49 [==============================] - 14s 297ms/step - loss: 0.0014 - acc: 1.0000 - val_loss: 0.7613 - val_acc: 0.8808\n",
            "Epoch 9/20\n",
            "49/49 [==============================] - 15s 301ms/step - loss: 8.8137e-05 - acc: 1.0000 - val_loss: 0.9491 - val_acc: 0.8806\n",
            "Epoch 10/20\n",
            "49/49 [==============================] - 14s 295ms/step - loss: 1.3070e-05 - acc: 1.0000 - val_loss: 1.1188 - val_acc: 0.8797\n",
            "Epoch 11/20\n",
            "49/49 [==============================] - 15s 304ms/step - loss: 1.7866e-06 - acc: 1.0000 - val_loss: 1.2795 - val_acc: 0.8798\n",
            "Epoch 12/20\n",
            "49/49 [==============================] - 15s 300ms/step - loss: 2.8556e-07 - acc: 1.0000 - val_loss: 1.4254 - val_acc: 0.8798\n",
            "Epoch 13/20\n",
            "49/49 [==============================] - 14s 297ms/step - loss: 6.5790e-08 - acc: 1.0000 - val_loss: 1.5329 - val_acc: 0.8795\n",
            "Epoch 14/20\n",
            "49/49 [==============================] - 15s 299ms/step - loss: 2.6372e-08 - acc: 1.0000 - val_loss: 1.5916 - val_acc: 0.8800\n",
            "Epoch 15/20\n",
            "49/49 [==============================] - 14s 293ms/step - loss: 1.4890e-08 - acc: 1.0000 - val_loss: 1.6255 - val_acc: 0.8795\n",
            "Epoch 16/20\n",
            "49/49 [==============================] - 14s 290ms/step - loss: 1.0642e-08 - acc: 1.0000 - val_loss: 1.6503 - val_acc: 0.8803\n",
            "Epoch 17/20\n",
            "49/49 [==============================] - 14s 292ms/step - loss: 8.0858e-09 - acc: 1.0000 - val_loss: 1.6710 - val_acc: 0.8801\n",
            "Epoch 18/20\n",
            "49/49 [==============================] - 15s 301ms/step - loss: 6.2834e-09 - acc: 1.0000 - val_loss: 1.6863 - val_acc: 0.8803\n",
            "Epoch 19/20\n",
            "49/49 [==============================] - 14s 290ms/step - loss: 5.6990e-09 - acc: 1.0000 - val_loss: 1.6987 - val_acc: 0.8803\n",
            "Epoch 20/20\n",
            "49/49 [==============================] - 15s 298ms/step - loss: 4.9189e-09 - acc: 1.0000 - val_loss: 1.7107 - val_acc: 0.8799\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5jzcwNgYnjES"
      },
      "source": [
        "## Exercise 11"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ozxFOw6hnlsl",
        "outputId": "e9942b19-2fb9-43f9-ed51-4771c697d9e3"
      },
      "source": [
        "original_train_loss = original_hist.history['loss']\n",
        "bigger_model_train_loss = bigger_hist.history['loss']\n",
        "\n",
        "plt.plot(epochs, original_train_loss, 'b+', label='Original model')\n",
        "plt.plot(epochs, bigger_model_train_loss, 'bo', label='Bigger model')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Training loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qArMB7Qln1SU"
      },
      "source": [
        "## Exercise 12"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8G0xdqNgoA6Y"
      },
      "source": [
        "l2_model = keras.Sequential([\n",
        "    keras.layers.Dense(16, activation='relu', input_shape=(10000,), kernel_regularizer = keras.regularizers.l2(0.001)),\n",
        "    keras.layers.Dense(16, activation='relu', kernel_regularizer = keras.regularizers.l2(0.001)),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "l2_model.compile(optimizer = 'rmsprop',\n",
        "                       loss = 'binary_crossentropy',\n",
        "                       metrics = ['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fcDXvqleoRhv"
      },
      "source": [
        "## Exercise 13"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2_xujzT5oS43",
        "outputId": "5a6aa8df-031c-4f1c-dd52-6d92e830f74c"
      },
      "source": [
        "l2_model_hist = l2_model.fit(x_train, y_train,\n",
        "                             epochs=20,\n",
        "                             batch_size=512,\n",
        "                             validation_data=(x_test, y_test))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "49/49 [==============================] - 3s 40ms/step - loss: 0.5991 - acc: 0.7562 - val_loss: 0.4053 - val_acc: 0.8797\n",
            "Epoch 2/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.3430 - acc: 0.9074 - val_loss: 0.3762 - val_acc: 0.8664\n",
            "Epoch 3/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.2832 - acc: 0.9198 - val_loss: 0.3360 - val_acc: 0.8878\n",
            "Epoch 4/20\n",
            "49/49 [==============================] - 1s 27ms/step - loss: 0.2522 - acc: 0.9316 - val_loss: 0.3384 - val_acc: 0.8852\n",
            "Epoch 5/20\n",
            "49/49 [==============================] - 1s 27ms/step - loss: 0.2426 - acc: 0.9344 - val_loss: 0.3452 - val_acc: 0.8828\n",
            "Epoch 6/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.2367 - acc: 0.9350 - val_loss: 0.3516 - val_acc: 0.8804\n",
            "Epoch 7/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.2280 - acc: 0.9399 - val_loss: 0.3703 - val_acc: 0.8760\n",
            "Epoch 8/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.2182 - acc: 0.9431 - val_loss: 0.3624 - val_acc: 0.8800\n",
            "Epoch 9/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.2135 - acc: 0.9457 - val_loss: 0.3845 - val_acc: 0.8732\n",
            "Epoch 10/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.2096 - acc: 0.9472 - val_loss: 0.4638 - val_acc: 0.8452\n",
            "Epoch 11/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.2151 - acc: 0.9440 - val_loss: 0.3901 - val_acc: 0.8715\n",
            "Epoch 12/20\n",
            "49/49 [==============================] - 1s 28ms/step - loss: 0.1985 - acc: 0.9534 - val_loss: 0.3840 - val_acc: 0.8758\n",
            "Epoch 13/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1988 - acc: 0.9526 - val_loss: 0.3887 - val_acc: 0.8729\n",
            "Epoch 14/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1973 - acc: 0.9516 - val_loss: 0.4039 - val_acc: 0.8715\n",
            "Epoch 15/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1926 - acc: 0.9535 - val_loss: 0.4090 - val_acc: 0.8692\n",
            "Epoch 16/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1981 - acc: 0.9483 - val_loss: 0.4072 - val_acc: 0.8709\n",
            "Epoch 17/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1927 - acc: 0.9531 - val_loss: 0.4452 - val_acc: 0.8626\n",
            "Epoch 18/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1967 - acc: 0.9500 - val_loss: 0.4117 - val_acc: 0.8711\n",
            "Epoch 19/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1903 - acc: 0.9530 - val_loss: 0.4235 - val_acc: 0.8653\n",
            "Epoch 20/20\n",
            "49/49 [==============================] - 1s 26ms/step - loss: 0.1883 - acc: 0.9541 - val_loss: 0.4477 - val_acc: 0.8631\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p3FqcEWSogvK"
      },
      "source": [
        "## Exercise 14"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yTMIX_OTordE",
        "outputId": "ab442b4d-e5d4-4cc8-aaba-432b57a45d8d"
      },
      "source": [
        "l2_model_val_loss = l2_model_hist.history['val_loss']\n",
        "\n",
        "plt.plot(epochs, original_val_loss, 'b+', label='Original model')\n",
        "plt.plot(epochs, l2_model_val_loss, 'bo', label='L2-regularized model')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Validation loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hKlPpo9fo-R9"
      },
      "source": [
        "We can see that the model with L2 regularization (dots) has become much more resistant to overfitting than the reference model (crosses), \n",
        "even though both models have the same number of parameters."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KNxt8aM8pFd4"
      },
      "source": [
        "## Exercise 15"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4jfzIjV5pWEY"
      },
      "source": [
        "dpt_model = keras.Sequential([\n",
        "    keras.layers.Dense(16, activation='relu', input_shape=(10000,)),\n",
        "    keras.layers.Dropout(0.5),\n",
        "    keras.layers.Dense(16, activation='relu'),\n",
        "    keras.layers.Dropout(0.5),\n",
        "    keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "dpt_model.compile(optimizer='rmsprop',\n",
        "                  loss='binary_crossentropy',\n",
        "                  metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ytnwarzmpbf4"
      },
      "source": [
        "## Exercise 16"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tQ2ULP-Ypc8c",
        "outputId": "7a9a039c-22a3-427f-a7b4-e5da37c46d24"
      },
      "source": [
        "dpt_hist = dpt_model.fit(x_train, y_train,\n",
        "                               epochs=20,\n",
        "                               batch_size=512,\n",
        "                               validation_data=(x_test, y_test))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "49/49 [==============================] - 3s 34ms/step - loss: 0.6404 - acc: 0.6176 - val_loss: 0.4469 - val_acc: 0.8656\n",
            "Epoch 2/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.4659 - acc: 0.7961 - val_loss: 0.3377 - val_acc: 0.8788\n",
            "Epoch 3/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.3750 - acc: 0.8528 - val_loss: 0.2944 - val_acc: 0.8889\n",
            "Epoch 4/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.3063 - acc: 0.8900 - val_loss: 0.2774 - val_acc: 0.8909\n",
            "Epoch 5/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.2713 - acc: 0.9088 - val_loss: 0.2795 - val_acc: 0.8893\n",
            "Epoch 6/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.2396 - acc: 0.9196 - val_loss: 0.3105 - val_acc: 0.8814\n",
            "Epoch 7/20\n",
            "49/49 [==============================] - 2s 32ms/step - loss: 0.2180 - acc: 0.9268 - val_loss: 0.2957 - val_acc: 0.8874\n",
            "Epoch 8/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.2017 - acc: 0.9349 - val_loss: 0.3194 - val_acc: 0.8861\n",
            "Epoch 9/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.1819 - acc: 0.9362 - val_loss: 0.3348 - val_acc: 0.8849\n",
            "Epoch 10/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.1691 - acc: 0.9434 - val_loss: 0.3685 - val_acc: 0.8817\n",
            "Epoch 11/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.1605 - acc: 0.9456 - val_loss: 0.4007 - val_acc: 0.8792\n",
            "Epoch 12/20\n",
            "49/49 [==============================] - 2s 32ms/step - loss: 0.1550 - acc: 0.9495 - val_loss: 0.4102 - val_acc: 0.8801\n",
            "Epoch 13/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.1402 - acc: 0.9559 - val_loss: 0.4427 - val_acc: 0.8762\n",
            "Epoch 14/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.1414 - acc: 0.9531 - val_loss: 0.4341 - val_acc: 0.8771\n",
            "Epoch 15/20\n",
            "49/49 [==============================] - 2s 32ms/step - loss: 0.1312 - acc: 0.9600 - val_loss: 0.4798 - val_acc: 0.8780\n",
            "Epoch 16/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.1259 - acc: 0.9578 - val_loss: 0.4732 - val_acc: 0.8746\n",
            "Epoch 17/20\n",
            "49/49 [==============================] - 1s 31ms/step - loss: 0.1368 - acc: 0.9561 - val_loss: 0.5024 - val_acc: 0.8757\n",
            "Epoch 18/20\n",
            "49/49 [==============================] - 2s 32ms/step - loss: 0.1240 - acc: 0.9593 - val_loss: 0.5378 - val_acc: 0.8776\n",
            "Epoch 19/20\n",
            "49/49 [==============================] - 2s 31ms/step - loss: 0.1190 - acc: 0.9593 - val_loss: 0.5358 - val_acc: 0.8748\n",
            "Epoch 20/20\n",
            "49/49 [==============================] - 1s 30ms/step - loss: 0.1247 - acc: 0.9600 - val_loss: 0.5665 - val_acc: 0.8746\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UtDnBWORps9z"
      },
      "source": [
        "## Exercise 17"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "crP28mF7puRw",
        "outputId": "c0d5df68-e0f6-42f4-fe2b-9f402fc1fdff"
      },
      "source": [
        "dpt_model_val_loss = dpt_hist.history['val_loss']\n",
        "\n",
        "plt.plot(epochs, original_val_loss, 'b+', label='Original model')\n",
        "plt.plot(epochs, dpt_model_val_loss, 'bo', label='Dropout-regularized model')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Validation loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eRgymbL5qD5-"
      },
      "source": [
        "We can see a clear improvement with this new model.  \n",
        "We conclude that the most common ways to prevent overfitting in neural networks are the following:\n",
        "- Getting more training data.\n",
        "- Reducing the capacity of the network.\n",
        "- Adding weight regularization.\n",
        "- Adding dropout."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TDReOKAHqh60"
      },
      "source": [
        "# Part 2"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "a-0_NabnqjFk"
      },
      "source": [
        "## Exercise 1"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Yo6t4TQ2hAag",
        "outputId": "e4d030ee-cee4-45cb-d7ac-bc37dc144abe"
      },
      "source": [
        "try :\n",
        "# %tensorflow_version only exists in Colab.\n",
        "  %tensorflow_version 2.x\n",
        "except Exception :\n",
        "  pass\n",
        "from keras . datasets import boston_housing\n",
        "\n",
        "(train_data, train_targets), (test_data, test_targets) = boston_housing.load_data()\n",
        "print(\"The train_data size is\", len(train_data), \"and the test_data size is\", len(test_data))\n",
        "print(\"The train_data shape is\", train_data.shape, \"and the test_data shape is\", test_data.shape)\n",
        "print(\"Here is the median value of the first owner-occupied homes (in thousands of dollars) :\", train_targets[0])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "The train_data size is 404 and the test_data size is 102\n",
            "The train_data shape is (404, 13) and the test_data shape is (102, 13)\n",
            "Here is the median value of the first owner-occupied homes (in thousands of dollars) : 15.2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wovJb02iiHfW"
      },
      "source": [
        "## Exercise 2"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aPx_4gadiMxn"
      },
      "source": [
        "mean = train_data.mean(axis=0)\n",
        "train_data -= mean\n",
        "test_data -= mean\n",
        "std = train_data.std(axis=0)\n",
        "train_data /= std\n",
        "test_data /= std"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cmQA-KSiizM1"
      },
      "source": [
        "In this way the feature is centered around 0 and has a unit standard deviation."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fVCR9Ux8itwe"
      },
      "source": [
        "## Exercise 3"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1TthHcbMi1lQ"
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from keras import layers\n",
        "from keras import models\n",
        "\n",
        "def build_model():\n",
        "  model = keras.Sequential([\n",
        "    keras.layers.Dense(64, activation='relu', input_shape=(train_data.shape[1],)),\n",
        "    keras.layers.Dense(64, activation='relu'),\n",
        "    keras.layers.Dense(1)\n",
        "])\n",
        "  model.compile(optimizer='rmsprop', loss='mse', metrics=['mae'])\n",
        "  return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0-vD1ngdjR2H"
      },
      "source": [
        "## Exercise 4"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Qr1su1rijTDH",
        "outputId": "db3da333-a0b8-4c72-a66c-c1aa605042c9"
      },
      "source": [
        "import numpy as np\n",
        "k = 4\n",
        "num_val_samples = len(train_data) // k\n",
        "num_epochs = 100\n",
        "all_scores = []\n",
        "for i in range(k):\n",
        "  print('processing fold #', i)\n",
        "  val_data = train_data[i * num_val_samples: (i + 1) * num_val_samples]\n",
        "  val_targets = train_targets[i * num_val_samples: (i + 1) * num_val_samples]\n",
        "  partial_train_data = np.concatenate([train_data[:i * num_val_samples], train_data[(i + 1) * num_val_samples:]], axis=0)\n",
        "  partial_train_targets = np.concatenate([train_targets[:i * num_val_samples], train_targets[(i + 1) * num_val_samples:]], axis=0)\n",
        "  model = build_model()\n",
        "  model.fit(partial_train_data, partial_train_targets, epochs=num_epochs, batch_size=1, verbose=0)\n",
        "  val_mse, val_mae = model.evaluate(val_data, val_targets, verbose=1) \n",
        "  all_scores.append(val_mae)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "processing fold # 0\n",
            "4/4 [==============================] - 0s 3ms/step - loss: 8.4824 - mae: 1.9388\n",
            "processing fold # 1\n",
            "4/4 [==============================] - 0s 2ms/step - loss: 31.4877 - mae: 3.7097\n",
            "processing fold # 2\n",
            "4/4 [==============================] - 0s 2ms/step - loss: 17.4626 - mae: 2.9804\n",
            "processing fold # 3\n",
            "4/4 [==============================] - 0s 2ms/step - loss: 11.9999 - mae: 2.4941\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tNIqOVHAkSLt"
      },
      "source": [
        "## Exercise 5"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jp1e2Jy8kTb2",
        "outputId": "b464acf6-aaf3-47a9-ed84-75a874d5847a"
      },
      "source": [
        "print(all_scores)\n",
        "np.mean(all_scores)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1.9388014078140259, 3.709718942642212, 2.9804084300994873, 2.4941015243530273]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "2.780757576227188"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xl8U0KkPkYgv"
      },
      "source": [
        "## Exercise 6"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "o-k3LvgMkZwj",
        "outputId": "11b469cb-c397-43b6-c400-2dfe704723b8"
      },
      "source": [
        "num_epochs = 500\n",
        "all_mae_history_train = []\n",
        "all_mae_histories = []\n",
        "for i in range(k):\n",
        "  print('processing fold #', i)\n",
        "  val_data = train_data[i * num_val_samples: (i + 1) * num_val_samples]\n",
        "  val_targets = train_targets[i * num_val_samples: (i + 1) * num_val_samples]\n",
        "  partial_train_data = np.concatenate([train_data[:i * num_val_samples], train_data[(i + 1) * num_val_samples:]], axis=0)\n",
        "  partial_train_targets = np.concatenate([train_targets[:i * num_val_samples], train_targets[(i + 1) * num_val_samples:]], axis=0)\n",
        "  model = build_model()\n",
        "  history = model.fit(partial_train_data, partial_train_targets, validation_data=(val_data, val_targets), epochs=num_epochs, batch_size=1, verbose=0)\n",
        "  mae_history = history.history['val_mae']\n",
        "  mae_history_train = history.history['mae']\n",
        "  print(mae_history_train)\n",
        "  all_mae_history_train.append(mae_history_train)\n",
        "  all_mae_histories.append(mae_history)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "processing fold # 0\n",
            "[10.428897857666016, 3.7943179607391357, 3.2296698093414307, 2.815258264541626, 2.62591814994812, 2.592055082321167, 2.5357367992401123, 2.439985513687134, 2.3653812408447266, 2.3792717456817627, 2.2804348468780518, 2.2352287769317627, 2.25079345703125, 2.17292857170105, 2.182953357696533, 2.1621270179748535, 2.107107639312744, 2.0728814601898193, 2.1202890872955322, 2.0593061447143555, 2.0743935108184814, 1.9904311895370483, 2.003239870071411, 1.9540385007858276, 2.0076534748077393, 2.029946804046631, 1.955426812171936, 2.0254015922546387, 1.9491301774978638, 1.9283334016799927, 1.8916795253753662, 1.8443697690963745, 1.8629693984985352, 1.915870189666748, 1.8419638872146606, 1.8493964672088623, 1.7742140293121338, 1.862583875656128, 1.852779746055603, 1.8492729663848877, 1.782726764678955, 1.7793833017349243, 1.8042281866073608, 1.7842669486999512, 1.8263825178146362, 1.7856816053390503, 1.7639540433883667, 1.8165124654769897, 1.7721434831619263, 1.691203236579895, 1.7032079696655273, 1.7226463556289673, 1.751575231552124, 1.7164907455444336, 1.715358853340149, 1.6968016624450684, 1.6760841608047485, 1.5704121589660645, 1.748014211654663, 1.6312787532806396, 1.62808096408844, 1.693996548652649, 1.6173958778381348, 1.6555452346801758, 1.604614019393921, 1.615060806274414, 1.6085004806518555, 1.6462395191192627, 1.6405746936798096, 1.577662467956543, 1.5668675899505615, 1.5365474224090576, 1.5899356603622437, 1.5670653581619263, 1.5436910390853882, 1.4731467962265015, 1.5342878103256226, 1.5214462280273438, 1.5210011005401611, 1.5564818382263184, 1.461958885192871, 1.536064624786377, 1.5181838274002075, 1.441444754600525, 1.5021334886550903, 1.4571497440338135, 1.4479467868804932, 1.486207365989685, 1.4123085737228394, 1.41896653175354, 1.4091180562973022, 1.4212403297424316, 1.43986976146698, 1.4035604000091553, 1.4233466386795044, 1.4015885591506958, 1.4000052213668823, 1.421417474746704, 1.3784242868423462, 1.3616368770599365, 1.376872181892395, 1.3367007970809937, 1.4040913581848145, 1.3043686151504517, 1.3728985786437988, 1.3799246549606323, 1.3813209533691406, 1.369289755821228, 1.2914949655532837, 1.3203853368759155, 1.3981170654296875, 1.3340685367584229, 1.3230922222137451, 1.2293139696121216, 1.3639742136001587, 1.2835710048675537, 1.3410861492156982, 1.289446473121643, 1.318835973739624, 1.2643173933029175, 1.2617132663726807, 1.2742795944213867, 1.3304595947265625, 1.2937638759613037, 1.289138913154602, 1.2206463813781738, 1.310295820236206, 1.2830215692520142, 1.2853608131408691, 1.309225082397461, 1.3003780841827393, 1.2397419214248657, 1.2619391679763794, 1.2326381206512451, 1.2423754930496216, 1.2485549449920654, 1.2288601398468018, 1.2014867067337036, 1.279341220855713, 1.2169933319091797, 1.2370120286941528, 1.252923607826233, 1.2715989351272583, 1.1673805713653564, 1.2025240659713745, 1.2190957069396973, 1.223992109298706, 1.213503360748291, 1.219436526298523, 1.2166290283203125, 1.2563401460647583, 1.2793192863464355, 1.1876965761184692, 1.2441818714141846, 1.148736834526062, 1.177445888519287, 1.1656177043914795, 1.1706560850143433, 1.188707709312439, 1.1781386137008667, 1.1317375898361206, 1.1981542110443115, 1.1643078327178955, 1.175614833831787, 1.1709184646606445, 1.1852322816848755, 1.0927187204360962, 1.1990211009979248, 1.1722285747528076, 1.1567680835723877, 1.1402257680892944, 1.1234544515609741, 1.1594130992889404, 1.1522278785705566, 1.153278112411499, 1.1225666999816895, 1.1610581874847412, 1.1105639934539795, 1.1466230154037476, 1.1083611249923706, 1.1810208559036255, 1.0884983539581299, 1.1257054805755615, 1.1135880947113037, 1.0724211931228638, 1.0674198865890503, 1.0755224227905273, 1.1076579093933105, 1.1142375469207764, 1.1126888990402222, 1.0927278995513916, 1.0847575664520264, 1.1331524848937988, 1.1012943983078003, 1.0769187211990356, 1.1390974521636963, 1.1036297082901, 1.0493286848068237, 1.0841453075408936, 1.0939818620681763, 1.074318289756775, 1.0573242902755737, 1.1047252416610718, 1.0625749826431274, 1.0611627101898193, 1.111134648323059, 1.032989501953125, 1.1019372940063477, 1.0937061309814453, 1.0216124057769775, 1.0036677122116089, 1.052577018737793, 1.091255784034729, 1.0608506202697754, 1.0601965188980103, 1.0394320487976074, 1.027340292930603, 1.0111291408538818, 1.0048941373825073, 1.0888911485671997, 1.0483272075653076, 1.0859133005142212, 1.0327985286712646, 1.039817214012146, 1.0493251085281372, 1.0431667566299438, 1.046291708946228, 1.0344526767730713, 1.0566916465759277, 1.0364022254943848, 1.0299973487854004, 1.059277057647705, 1.032469630241394, 1.0305842161178589, 1.0217205286026, 1.018985629081726, 0.9871275424957275, 1.0493956804275513, 1.0033756494522095, 1.0508607625961304, 1.006946325302124, 1.066956639289856, 0.9483098983764648, 1.0194402933120728, 1.0145131349563599, 0.9927030801773071, 0.972864031791687, 1.0022708177566528, 1.0052220821380615, 0.9872347712516785, 1.0187734365463257, 0.9922747015953064, 0.9751756191253662, 0.9616675972938538, 1.0035650730133057, 0.9539889693260193, 0.8954442739486694, 1.0442901849746704, 1.0095927715301514, 0.9496280550956726, 0.9659821391105652, 0.9379076957702637, 0.9462427496910095, 0.9865464568138123, 0.9488224387168884, 0.9905088543891907, 0.9461630582809448, 0.9379686117172241, 0.9887016415596008, 0.9846771359443665, 0.9714881181716919, 0.9347439408302307, 0.9739001393318176, 0.9387127757072449, 0.9320664405822754, 0.9030142426490784, 0.8986604809761047, 0.9075750708580017, 0.8918679356575012, 0.9067932963371277, 0.9389533400535583, 0.9327009916305542, 0.8969315886497498, 0.9320048093795776, 0.8916404247283936, 0.9009812474250793, 0.9654701948165894, 0.9079594016075134, 0.9306504726409912, 0.9373539686203003, 0.8847813606262207, 0.9117137789726257, 0.9022263288497925, 0.8804734945297241, 0.9165735244750977, 0.89247065782547, 0.9221433401107788, 0.8709861040115356, 0.8764419555664062, 0.928600549697876, 0.8916215896606445, 0.8797541856765747, 0.8747069835662842, 0.84123694896698, 0.8556809425354004, 0.9651912450790405, 0.8875046372413635, 0.8870931267738342, 0.8605730533599854, 0.8982800841331482, 0.8433853387832642, 0.8885465860366821, 0.9150568842887878, 0.8559885621070862, 0.8687938451766968, 0.9275721907615662, 0.9147198796272278, 0.8364827632904053, 0.854880154132843, 0.8861965537071228, 0.8758016228675842, 0.9034652709960938, 0.8762538433074951, 0.8323084712028503, 0.8776258230209351, 0.8687700033187866, 0.8586797714233398, 0.8582801818847656, 0.8485307097434998, 0.8562440872192383, 0.8072029948234558, 0.8292344808578491, 0.831699550151825, 0.8829675912857056, 0.8114802837371826, 0.8779985904693604, 0.8683175444602966, 0.7697046995162964, 0.8349267840385437, 0.8533143997192383, 0.8368935585021973, 0.8134505152702332, 0.8509302139282227, 0.8298487067222595, 0.8541823625564575, 0.8261110186576843, 0.8307815790176392, 0.7945122718811035, 0.799560546875, 0.8440893292427063, 0.8291016817092896, 0.8524726033210754, 0.830965518951416, 0.8328489065170288, 0.7916069626808167, 0.854168176651001, 0.8036068081855774, 0.7847252488136292, 0.7979211211204529, 0.8218885064125061, 0.8562639355659485, 0.8137133717536926, 0.8257095813751221, 0.8314971923828125, 0.8227034211158752, 0.7926949262619019, 0.8124316334724426, 0.8184928297996521, 0.7615220546722412, 0.7901430130004883, 0.8080974817276001, 0.7839775681495667, 0.815442681312561, 0.8079286813735962, 0.8143550753593445, 0.7542377710342407, 0.7659137845039368, 0.8009470105171204, 0.8130248188972473, 0.7513601183891296, 0.8637089133262634, 0.7414870858192444, 0.8121901750564575, 0.7686120271682739, 0.789953351020813, 0.7621984481811523, 0.7826599478721619, 0.7377902865409851, 0.8658072352409363, 0.7673582434654236, 0.7431294918060303, 0.7612882852554321, 0.7206360697746277, 0.7878444194793701, 0.735262393951416, 0.7581340670585632, 0.7908287048339844, 0.8131400346755981, 0.7901040315628052, 0.7704113125801086, 0.827947735786438, 0.7963180541992188, 0.782468318939209, 0.7568178176879883, 0.7869310975074768, 0.7754790782928467, 0.7661823034286499, 0.7335518598556519, 0.7615431547164917, 0.7216469645500183, 0.7411039471626282, 0.7332378029823303, 0.8115267157554626, 0.7852737903594971, 0.7425351142883301, 0.738544225692749, 0.7591060996055603, 0.737865149974823, 0.8015487194061279, 0.7159868478775024, 0.7553428411483765, 0.7866949439048767, 0.7750560641288757, 0.7322130799293518, 0.692134439945221, 0.6924752593040466, 0.7623547911643982, 0.7273262739181519, 0.7295949459075928, 0.763504683971405, 0.7314961552619934, 0.7619694471359253, 0.7379204630851746, 0.7594023942947388, 0.7449114918708801, 0.7641528248786926, 0.7392247915267944, 0.6990063190460205, 0.7068741321563721, 0.7899585962295532, 0.706591010093689, 0.7464832067489624, 0.7222246527671814, 0.7684076428413391, 0.7749898433685303, 0.7514525651931763, 0.7093465328216553, 0.751098096370697, 0.7221572399139404, 0.7405619621276855, 0.7148931622505188, 0.7263376116752625, 0.7218571305274963, 0.6871942281723022, 0.7551971673965454, 0.7274610996246338, 0.7274672985076904, 0.7365282773971558, 0.7110806107521057, 0.7072075009346008, 0.707241952419281, 0.7356845140457153, 0.6899226903915405, 0.7406880855560303, 0.6682581901550293, 0.7183204889297485, 0.7004867196083069, 0.7186813354492188, 0.7379598021507263, 0.7298767566680908, 0.7406588792800903, 0.691950798034668, 0.7093064785003662, 0.7149072289466858, 0.6840662360191345, 0.7335323691368103, 0.6687906384468079, 0.7055090665817261, 0.7244065403938293, 0.6956456303596497, 0.7051896452903748, 0.6929857730865479, 0.6980280876159668, 0.6909066438674927, 0.6631380915641785, 0.6986331939697266, 0.7004698514938354, 0.7334032654762268, 0.6996445655822754, 0.6893843412399292, 0.7030454874038696, 0.6890507340431213, 0.6861785650253296, 0.6787105202674866, 0.6480352282524109, 0.6939213871955872, 0.7040907144546509, 0.696018397808075, 0.6527659893035889, 0.6767459511756897]\n",
            "processing fold # 1\n",
            "[10.725848197937012, 3.6612319946289062, 3.0213959217071533, 2.762484312057495, 2.4996750354766846, 2.4347047805786133, 2.298835277557373, 2.2433619499206543, 2.1346306800842285, 2.145981788635254, 2.132838249206543, 2.0918421745300293, 2.0733375549316406, 2.038109540939331, 2.0687789916992188, 2.0437910556793213, 1.9440374374389648, 1.9272797107696533, 1.9995900392532349, 1.9733121395111084, 1.9649953842163086, 1.9249290227890015, 1.87582266330719, 1.9426108598709106, 1.8713136911392212, 1.9518364667892456, 1.9287302494049072, 1.8991448879241943, 1.8654919862747192, 1.8755724430084229, 1.812930703163147, 1.8693389892578125, 1.7803165912628174, 1.8272594213485718, 1.7743091583251953, 1.7519773244857788, 1.798793911933899, 1.792834758758545, 1.757172703742981, 1.7943397760391235, 1.7730549573898315, 1.7514417171478271, 1.7187026739120483, 1.6983370780944824, 1.7124567031860352, 1.7340257167816162, 1.7555139064788818, 1.7591713666915894, 1.7528014183044434, 1.7046570777893066, 1.6908364295959473, 1.6725473403930664, 1.5791760683059692, 1.6732388734817505, 1.687867522239685, 1.649611234664917, 1.6279815435409546, 1.6192080974578857, 1.6145321130752563, 1.6032146215438843, 1.6098374128341675, 1.526890516281128, 1.536574363708496, 1.5513023138046265, 1.6037700176239014, 1.6085065603256226, 1.5282657146453857, 1.5772335529327393, 1.5513144731521606, 1.5455307960510254, 1.4897304773330688, 1.566872239112854, 1.4301841259002686, 1.5062618255615234, 1.4712742567062378, 1.5100654363632202, 1.4726274013519287, 1.4397811889648438, 1.4782841205596924, 1.4813644886016846, 1.385855793952942, 1.400173306465149, 1.3664891719818115, 1.4156802892684937, 1.434382438659668, 1.3549587726593018, 1.36451256275177, 1.379272222518921, 1.3837109804153442, 1.3516143560409546, 1.3656762838363647, 1.3705483675003052, 1.3712486028671265, 1.3568036556243896, 1.3780711889266968, 1.2999557256698608, 1.2987382411956787, 1.2892717123031616, 1.279511570930481, 1.2923434972763062, 1.2716073989868164, 1.2832280397415161, 1.2230565547943115, 1.3014718294143677, 1.308095932006836, 1.2788175344467163, 1.2648019790649414, 1.2408488988876343, 1.2728705406188965, 1.2172006368637085, 1.1696715354919434, 1.255106806755066, 1.2429555654525757, 1.1796901226043701, 1.2224202156066895, 1.2627499103546143, 1.1886519193649292, 1.1993696689605713, 1.215264081954956, 1.2346221208572388, 1.2024571895599365, 1.2086859941482544, 1.1305183172225952, 1.150498628616333, 1.1690998077392578, 1.1741199493408203, 1.1726254224777222, 1.1577599048614502, 1.1200430393218994, 1.2266522645950317, 1.100701928138733, 1.1695035696029663, 1.1265738010406494, 1.1668787002563477, 1.1224230527877808, 1.086181402206421, 1.1165752410888672, 1.149153232574463, 1.0924582481384277, 1.136459469795227, 1.1439100503921509, 1.0757359266281128, 1.1163029670715332, 1.0387513637542725, 1.0790349245071411, 1.10794198513031, 1.1054041385650635, 1.0054843425750732, 1.116889238357544, 1.066929578781128, 1.0879837274551392, 1.0647073984146118, 1.0202062129974365, 1.0210282802581787, 1.0950783491134644, 1.0572844743728638, 1.0301885604858398, 1.0813688039779663, 1.0230053663253784, 1.0477343797683716, 1.0321545600891113, 1.0541422367095947, 1.0638407468795776, 1.103449821472168, 1.0668803453445435, 1.0067068338394165, 0.9935511350631714, 1.0024925470352173, 0.9875665307044983, 1.0587143898010254, 1.0128819942474365, 1.0113697052001953, 0.9542718529701233, 0.9954847097396851, 1.042595386505127, 1.0046372413635254, 1.0426074266433716, 1.0458557605743408, 0.9822152853012085, 0.9968416690826416, 1.0304237604141235, 0.9890706539154053, 0.9459476470947266, 0.9782742857933044, 1.0071016550064087, 0.9902408123016357, 0.9253093004226685, 0.9239687323570251, 1.0087807178497314, 0.9599029421806335, 0.9782592058181763, 0.9624265432357788, 0.9532162547111511, 0.9444406628608704, 0.9189988970756531, 0.9130004048347473, 0.9695460796356201, 0.9208285212516785, 0.9129543900489807, 0.9240633249282837, 0.9212241768836975, 0.9734232425689697, 0.9214677214622498, 0.9296104311943054, 0.9090048670768738, 0.9226556420326233, 0.9538878202438354, 0.8991353511810303, 0.9061764478683472, 0.9000772833824158, 0.8970922827720642, 0.9503499865531921, 0.9223686456680298, 0.8971412777900696, 0.9113784432411194, 0.9272156953811646, 0.8675804138183594, 0.8982774615287781, 0.9097323417663574, 0.8716147541999817, 0.8967903852462769, 0.8437088131904602, 0.8884720206260681, 0.9037180542945862, 0.8514132499694824, 0.9151232838630676, 0.8705546855926514, 0.8329890966415405, 0.8650109767913818, 0.9034759402275085, 0.8971112370491028, 0.8694176077842712, 0.8838505148887634, 0.8191692233085632, 0.8601452708244324, 0.8460354208946228, 0.8380849957466125, 0.8841830492019653, 0.8656508326530457, 0.8519073724746704, 0.8344335556030273, 0.8259850144386292, 0.8290202617645264, 0.8501230478286743, 0.7961715459823608, 0.8292611241340637, 0.8076331615447998, 0.8102589845657349, 0.8179399967193604, 0.8217551112174988, 0.8333777785301208, 0.8609724044799805, 0.7897529006004333, 0.8239802122116089, 0.8686342239379883, 0.7957040071487427, 0.8392452001571655, 0.8303102850914001, 0.810454249382019, 0.8028412461280823, 0.8274775147438049, 0.8063960075378418, 0.8488919734954834, 0.798335611820221, 0.8219665884971619, 0.8332647085189819, 0.7815608382225037, 0.7773455381393433, 0.8062748312950134, 0.7959330081939697, 0.7911656498908997, 0.8433431386947632, 0.810407280921936, 0.7845407724380493, 0.8318371772766113, 0.8162342309951782, 0.7667710781097412, 0.779097318649292, 0.7519906163215637, 0.7706048488616943, 0.751431405544281, 0.7823653817176819, 0.812886655330658, 0.7747999429702759, 0.7715106010437012, 0.7439252138137817, 0.7301125526428223, 0.7979814410209656, 0.8176780939102173, 0.7957502603530884, 0.7060174345970154, 0.7813414335250854, 0.824510931968689, 0.7908197045326233, 0.7736935615539551, 0.7731701731681824, 0.7384182214736938, 0.8322299122810364, 0.7602106928825378, 0.739933431148529, 0.7703630328178406, 0.7358238697052002, 0.7303312420845032, 0.7268863916397095, 0.7830929160118103, 0.7446433305740356, 0.7601868510246277, 0.7154483199119568, 0.7756475210189819, 0.7409366369247437, 0.710124135017395, 0.7153500318527222, 0.7179978489875793, 0.7723599672317505, 0.7280418872833252, 0.7400994300842285, 0.7538019418716431, 0.729022741317749, 0.7554484605789185, 0.6674978137016296, 0.7456640601158142, 0.7576866149902344, 0.7236904501914978, 0.7171147465705872, 0.6997683644294739, 0.7248691916465759, 0.7454822659492493, 0.7314753532409668, 0.72060626745224, 0.6841577887535095, 0.7165700793266296, 0.7304902076721191, 0.7386700510978699, 0.6890098452568054, 0.7351155281066895, 0.7195147275924683, 0.711383044719696, 0.7088702321052551, 0.712386965751648, 0.7673832178115845, 0.6998317241668701, 0.7016265392303467, 0.736377477645874, 0.7016413807868958, 0.7469431161880493, 0.7023213505744934, 0.6964982748031616, 0.7041018009185791, 0.7440345287322998, 0.7168356776237488, 0.7286067605018616, 0.7159661650657654, 0.6943520307540894, 0.7161571979522705, 0.6848354935646057, 0.6804268956184387, 0.7116025686264038, 0.7126777768135071, 0.6618403792381287, 0.7036943435668945, 0.6840035319328308, 0.6794214248657227, 0.6782915592193604, 0.7146475315093994, 0.6956790685653687, 0.7193511128425598, 0.6792734265327454, 0.6705110669136047, 0.7080413699150085, 0.6712636351585388, 0.6836198568344116, 0.635466992855072, 0.6822471022605896, 0.6590914130210876, 0.6775771975517273, 0.7074074745178223, 0.6933616399765015, 0.6907402873039246, 0.6751276850700378, 0.6905477046966553, 0.6648113131523132, 0.6972859501838684, 0.684376060962677, 0.685886561870575, 0.690769374370575, 0.6883736252784729, 0.6756806373596191, 0.6799122095108032, 0.6785844564437866, 0.6635243892669678, 0.6820833086967468, 0.6494845747947693, 0.667580783367157, 0.6914855241775513, 0.6898064017295837, 0.6997216939926147, 0.6737809181213379, 0.6523410081863403, 0.659400463104248, 0.6839205622673035, 0.6585136651992798, 0.6439637541770935, 0.6667863130569458, 0.6696174740791321, 0.6654608845710754, 0.684502124786377, 0.660722553730011, 0.6301795244216919, 0.6998880505561829, 0.6691636443138123, 0.6721946597099304, 0.6376992464065552, 0.6378664374351501, 0.6866135001182556, 0.6356224417686462, 0.6964265704154968, 0.6444857716560364, 0.6405581831932068, 0.6564809083938599, 0.649105429649353, 0.641630232334137, 0.6499184370040894, 0.6169646978378296, 0.6786565780639648, 0.6326475739479065, 0.6747995615005493, 0.6241267323493958, 0.6711127758026123, 0.6602940559387207, 0.6695508360862732, 0.6372706294059753, 0.6507758498191833, 0.6579163074493408, 0.6244936585426331, 0.6506848931312561, 0.6326448917388916, 0.6421309113502502, 0.6475308537483215, 0.6355865001678467, 0.6648609638214111, 0.6435533761978149, 0.6279115080833435, 0.6665011644363403, 0.6109244227409363, 0.6343641877174377, 0.6540297865867615, 0.633326530456543, 0.6626473665237427, 0.6263808608055115, 0.682014524936676, 0.6055117845535278, 0.6966041326522827, 0.657416582107544, 0.5954504013061523, 0.6444331407546997, 0.623144268989563, 0.6470958590507507, 0.6572464108467102, 0.6351008415222168, 0.6213814616203308, 0.6550157070159912, 0.6465053558349609, 0.6288188099861145, 0.6119089722633362, 0.6100371479988098, 0.6555343270301819, 0.6319621205329895, 0.6413529515266418, 0.6102475523948669, 0.5959725379943848, 0.6213356852531433, 0.6500428915023804, 0.6087629199028015, 0.6319840550422668, 0.6374881863594055, 0.6060255169868469, 0.6122382879257202, 0.6086336374282837, 0.6456050872802734, 0.6215565204620361, 0.6379731297492981, 0.6126776933670044, 0.6053315997123718, 0.6202709078788757, 0.6226593255996704, 0.643882155418396, 0.6064508557319641, 0.6200244426727295, 0.6440258622169495, 0.6072262525558472, 0.6186190843582153, 0.6258663535118103, 0.5920493006706238, 0.6512789726257324, 0.5990053415298462, 0.6303837299346924, 0.5875859260559082, 0.6160786151885986, 0.59880131483078, 0.6161394715309143]\n",
            "processing fold # 2\n",
            "[11.279312133789062, 3.6465513706207275, 3.0009660720825195, 2.723876714706421, 2.573575258255005, 2.5215718746185303, 2.3150484561920166, 2.2840187549591064, 2.319091796875, 2.144287586212158, 2.2216036319732666, 2.144794464111328, 2.085327386856079, 2.021146059036255, 2.0122926235198975, 1.9836318492889404, 1.9829460382461548, 1.9100093841552734, 1.9290064573287964, 1.882304072380066, 1.8985612392425537, 1.8491462469100952, 1.862194538116455, 1.8011711835861206, 1.8172786235809326, 1.8131730556488037, 1.7118314504623413, 1.7491153478622437, 1.6886223554611206, 1.7308112382888794, 1.6849547624588013, 1.7014204263687134, 1.6585084199905396, 1.687015175819397, 1.6331676244735718, 1.6439793109893799, 1.6403908729553223, 1.6415510177612305, 1.6180269718170166, 1.48051118850708, 1.6074001789093018, 1.635505199432373, 1.568827509880066, 1.5707770586013794, 1.572221040725708, 1.5133066177368164, 1.5450646877288818, 1.4577627182006836, 1.5108106136322021, 1.4604250192642212, 1.459600567817688, 1.4741023778915405, 1.5058388710021973, 1.4828096628189087, 1.3943639993667603, 1.444714069366455, 1.379438042640686, 1.419830322265625, 1.3895761966705322, 1.4181327819824219, 1.421660304069519, 1.388953447341919, 1.398563027381897, 1.3528177738189697, 1.3829214572906494, 1.3254917860031128, 1.3671246767044067, 1.3866541385650635, 1.3704464435577393, 1.3993443250656128, 1.320729374885559, 1.2695918083190918, 1.3507252931594849, 1.2997435331344604, 1.3048690557479858, 1.2889037132263184, 1.2957935333251953, 1.3072630167007446, 1.3092467784881592, 1.2737475633621216, 1.3037388324737549, 1.275450348854065, 1.2541383504867554, 1.2791569232940674, 1.2347514629364014, 1.2470810413360596, 1.1960651874542236, 1.2526344060897827, 1.2001656293869019, 1.1820827722549438, 1.2541611194610596, 1.1571933031082153, 1.1765652894973755, 1.1995426416397095, 1.1493277549743652, 1.1610467433929443, 1.1771236658096313, 1.1876552104949951, 1.1821610927581787, 1.1536530256271362, 1.10647451877594, 1.1751489639282227, 1.1394098997116089, 1.2148176431655884, 1.1350092887878418, 1.1936951875686646, 1.1355619430541992, 1.1407915353775024, 1.1625292301177979, 1.1869490146636963, 1.105968713760376, 1.111893653869629, 1.152693271636963, 1.0599480867385864, 1.1176286935806274, 1.1017040014266968, 1.0983279943466187, 1.1239811182022095, 1.1149210929870605, 1.1076008081436157, 1.0497714281082153, 1.1349178552627563, 1.0211296081542969, 1.0522788763046265, 1.0077457427978516, 1.1345750093460083, 1.108568549156189, 1.0536152124404907, 1.0947139263153076, 1.0733153820037842, 1.0201120376586914, 1.0814071893692017, 1.0419259071350098, 1.082167625427246, 1.0167315006256104, 1.0506821870803833, 1.0528602600097656, 1.0384434461593628, 1.0475589036941528, 0.9948570132255554, 0.9812230467796326, 1.0058149099349976, 0.978075385093689, 1.0038764476776123, 0.9989821314811707, 1.0423076152801514, 1.0020363330841064, 0.9934234023094177, 0.9991703629493713, 0.9953154921531677, 0.9895967245101929, 0.9449836611747742, 0.9693467617034912, 1.0000932216644287, 0.9724576473236084, 0.9628947377204895, 0.9602185487747192, 0.950493335723877, 0.9617125391960144, 0.9564579725265503, 0.9036237597465515, 0.9569292068481445, 0.9215326309204102, 0.9329245686531067, 0.9148316979408264, 0.8649501204490662, 0.8550404906272888, 0.9296239614486694, 0.9639729261398315, 0.8456273078918457, 0.9345736503601074, 0.9380508065223694, 0.9439471364021301, 0.943619430065155, 0.9299202561378479, 0.9517223834991455, 0.874315619468689, 0.9017974734306335, 0.9084440469741821, 0.8955509066581726, 0.9137037396430969, 0.9328382611274719, 0.898215115070343, 0.8648363351821899, 0.9151992797851562, 0.9145296216011047, 0.9327690601348877, 0.9172616004943848, 0.8649806380271912, 0.8970965147018433, 0.8865541815757751, 0.8571768999099731, 0.8860647082328796, 0.8717175126075745, 0.8919525742530823, 0.8913156986236572, 0.8788928389549255, 0.8499273657798767, 0.8664754033088684, 0.886086642742157, 0.8445355296134949, 0.8418659567832947, 0.8697826266288757, 0.8450825810432434, 0.8690930604934692, 0.812845766544342, 0.8874796628952026, 0.8298566341400146, 0.8377341628074646, 0.8746778964996338, 0.8467181324958801, 0.8454306721687317, 0.8418347239494324, 0.8659126162528992, 0.8885363340377808, 0.798578143119812, 0.8121826648712158, 0.7936809659004211, 0.8398448824882507, 0.8414738774299622, 0.8444693088531494, 0.8415188789367676, 0.8311092257499695, 0.8769879937171936, 0.8564717769622803, 0.8913955092430115, 0.7972071766853333, 0.815720796585083, 0.8639209866523743, 0.8563463091850281, 0.8574652671813965, 0.8301432728767395, 0.7985110878944397, 0.8644837141036987, 0.8629242777824402, 0.7843000292778015, 0.8137577772140503, 0.7861846685409546, 0.8001932501792908, 0.826106607913971, 0.8047821521759033, 0.8257138729095459, 0.7757349610328674, 0.7942256331443787, 0.8562809228897095, 0.7648034691810608, 0.7964136600494385, 0.7968392372131348, 0.8002097010612488, 0.8009008169174194, 0.8058552145957947, 0.7866913080215454, 0.7451003193855286, 0.8207074999809265, 0.7635947465896606, 0.8465611338615417, 0.7920525670051575, 0.7811955213546753, 0.7805526256561279, 0.7525081038475037, 0.7811169028282166, 0.7553418278694153, 0.7794582843780518, 0.7422336935997009, 0.7530248761177063, 0.7784444093704224, 0.7813952565193176, 0.7237004637718201, 0.7855161428451538, 0.7887652516365051, 0.7676325440406799, 0.7285232543945312, 0.7556699514389038, 0.7961981892585754, 0.7843191623687744, 0.8031350374221802, 0.7207328081130981, 0.7408551573753357, 0.7536733150482178, 0.7627698183059692, 0.7260491251945496, 0.7657005786895752, 0.740273654460907, 0.7662030458450317, 0.7557728886604309, 0.7637807726860046, 0.7336422801017761, 0.7796068787574768, 0.7327668070793152, 0.7523257732391357, 0.7129090428352356, 0.7478071451187134, 0.749276876449585, 0.7142975926399231, 0.7147021293640137, 0.7773640751838684, 0.7578229904174805, 0.7456637024879456, 0.7461021542549133, 0.7247536778450012, 0.7492635250091553, 0.6976204514503479, 0.7499632835388184, 0.7128974795341492, 0.7186247110366821, 0.7128711342811584, 0.7679570317268372, 0.7176642417907715, 0.7038187384605408, 0.7371874451637268, 0.7154341340065002, 0.7233679890632629, 0.6929288506507874, 0.6851518750190735, 0.6885192394256592, 0.719634473323822, 0.7047957181930542, 0.7251766324043274, 0.6827065348625183, 0.709647536277771, 0.7109936475753784, 0.7010042071342468, 0.6893467903137207, 0.7133156657218933, 0.7248796820640564, 0.7055122256278992, 0.7534286379814148, 0.6974719166755676, 0.7359179258346558, 0.7022175788879395, 0.7131616473197937, 0.6741922497749329, 0.7221574783325195, 0.711735725402832, 0.7409834265708923, 0.6981510519981384, 0.6778556108474731, 0.691949188709259, 0.7332736253738403, 0.6881238222122192, 0.624540388584137, 0.6844907402992249, 0.7344652414321899, 0.6279572248458862, 0.7070663571357727, 0.6360880732536316, 0.6823055148124695, 0.7118369340896606, 0.6787744164466858, 0.6810865998268127, 0.702163815498352, 0.6320605874061584, 0.7118869423866272, 0.6563849449157715, 0.6892474889755249, 0.631367564201355, 0.6927961707115173, 0.6836516261100769, 0.6983461380004883, 0.6607069969177246, 0.6783946752548218, 0.6977661848068237, 0.6923341155052185, 0.6669543981552124, 0.6552350521087646, 0.6596431732177734, 0.7048391699790955, 0.6879226565361023, 0.6896609663963318, 0.6513209342956543, 0.6768317222595215, 0.6894591450691223, 0.6894459128379822, 0.6579620838165283, 0.6746206879615784, 0.6907219886779785, 0.722933292388916, 0.6704051494598389, 0.6485477685928345, 0.6478566527366638, 0.7013751268386841, 0.6588822603225708, 0.6304027438163757, 0.6691792607307434, 0.6999182105064392, 0.6939533352851868, 0.6485692858695984, 0.6391089558601379, 0.6586103439331055, 0.6426883935928345, 0.696872889995575, 0.6666863560676575, 0.6751449704170227, 0.636169970035553, 0.6609970331192017, 0.6441106200218201, 0.6908210515975952, 0.6417303085327148, 0.6895123720169067, 0.6461349129676819, 0.660527229309082, 0.6157312393188477, 0.6805760860443115, 0.6276147365570068, 0.6327409148216248, 0.6359250545501709, 0.6576665639877319, 0.6530776023864746, 0.6700606942176819, 0.6314919590950012, 0.6538981199264526, 0.6928816437721252, 0.6790230870246887, 0.6534377932548523, 0.6311728358268738, 0.6614347696304321, 0.6303955912590027, 0.6761044859886169, 0.6467710733413696, 0.6360555291175842, 0.5975702404975891, 0.6363850831985474, 0.6434106230735779, 0.631149411201477, 0.6497816443443298, 0.6283556818962097, 0.6359533667564392, 0.6684176325798035, 0.6133732795715332, 0.6436551213264465, 0.661271333694458, 0.6118602156639099, 0.6468095183372498, 0.6702925562858582, 0.6424422860145569, 0.6396239399909973, 0.6069371104240417, 0.6352793574333191, 0.6198880672454834, 0.6629604697227478, 0.6102027297019958, 0.6378253698348999, 0.6306281089782715, 0.617935836315155, 0.6205491423606873, 0.6376800537109375, 0.6274029612541199, 0.6537425518035889, 0.6320695877075195, 0.613960325717926, 0.6107676029205322, 0.6437736749649048, 0.6345677375793457, 0.6102384328842163, 0.6388945579528809, 0.6258765459060669, 0.5864626169204712, 0.6107012629508972, 0.6141948699951172, 0.5964863300323486, 0.633863627910614, 0.6230214238166809, 0.6021270155906677, 0.6485995650291443, 0.6090372204780579, 0.6105257272720337, 0.6147237420082092, 0.583598256111145, 0.6375507116317749, 0.6086537837982178, 0.5913915038108826, 0.6309139728546143, 0.6518632769584656, 0.592430591583252, 0.6100358366966248, 0.609337568283081, 0.6529350876808167, 0.5772456526756287, 0.6298946738243103, 0.6004582643508911, 0.6465855240821838, 0.614794135093689, 0.6171982884407043, 0.603280782699585, 0.6099913716316223, 0.598206102848053, 0.6242181062698364, 0.632497251033783, 0.5922158360481262, 0.6012324690818787, 0.5957624912261963, 0.6453271508216858, 0.6102734208106995, 0.5894424915313721, 0.6244598627090454, 0.6121727228164673, 0.5982618927955627, 0.6297467350959778, 0.611709475517273, 0.5995202660560608]\n",
            "processing fold # 3\n",
            "[10.306893348693848, 3.5212204456329346, 3.0301105976104736, 2.779860019683838, 2.559974193572998, 2.4760684967041016, 2.349343776702881, 2.3350651264190674, 2.2187938690185547, 2.156651020050049, 2.195746898651123, 2.177246332168579, 2.0320842266082764, 2.105569362640381, 2.01537823677063, 2.1136415004730225, 2.0307466983795166, 2.029763698577881, 1.9996607303619385, 2.036421537399292, 1.9554046392440796, 1.9420055150985718, 1.903936505317688, 1.8801928758621216, 1.922631025314331, 1.9116374254226685, 1.9199309349060059, 1.867748737335205, 1.8665119409561157, 1.8937822580337524, 1.9192209243774414, 1.8159503936767578, 1.8215712308883667, 1.8339107036590576, 1.8620319366455078, 1.7221448421478271, 1.8074915409088135, 1.7558233737945557, 1.7817898988723755, 1.7252048254013062, 1.747456669807434, 1.7592209577560425, 1.7464555501937866, 1.7442965507507324, 1.7567185163497925, 1.7150254249572754, 1.6764732599258423, 1.6814370155334473, 1.6170655488967896, 1.706075668334961, 1.6967511177062988, 1.6186097860336304, 1.643828272819519, 1.6461186408996582, 1.6657148599624634, 1.5982143878936768, 1.5756467580795288, 1.621084451675415, 1.5728745460510254, 1.5884242057800293, 1.5148190259933472, 1.6170247793197632, 1.5520049333572388, 1.5755832195281982, 1.567734956741333, 1.5312232971191406, 1.495124340057373, 1.537622094154358, 1.5076671838760376, 1.5336799621582031, 1.5080790519714355, 1.5006024837493896, 1.5047211647033691, 1.5037052631378174, 1.5046957731246948, 1.4446505308151245, 1.4292078018188477, 1.4295144081115723, 1.4430378675460815, 1.472121000289917, 1.4704816341400146, 1.4507511854171753, 1.3941826820373535, 1.411924958229065, 1.3902063369750977, 1.4109715223312378, 1.367518663406372, 1.401480793952942, 1.3629364967346191, 1.3772602081298828, 1.3613262176513672, 1.3489021062850952, 1.3544477224349976, 1.3592928647994995, 1.2864948511123657, 1.3771538734436035, 1.3471792936325073, 1.3323744535446167, 1.334120512008667, 1.3136262893676758, 1.3239655494689941, 1.2947033643722534, 1.3023494482040405, 1.2599918842315674, 1.3125110864639282, 1.2888867855072021, 1.2179285287857056, 1.3103572130203247, 1.2677717208862305, 1.2599537372589111, 1.249239206314087, 1.2717818021774292, 1.2232375144958496, 1.2734144926071167, 1.26063871383667, 1.230574369430542, 1.2436423301696777, 1.260902762413025, 1.2447532415390015, 1.2010233402252197, 1.1932506561279297, 1.237420916557312, 1.2327364683151245, 1.1780391931533813, 1.2119170427322388, 1.1513864994049072, 1.1679105758666992, 1.1564531326293945, 1.1836843490600586, 1.1792396306991577, 1.1701767444610596, 1.187058448791504, 1.1635940074920654, 1.1688742637634277, 1.1668314933776855, 1.1461530923843384, 1.1389936208724976, 1.1599512100219727, 1.1638094186782837, 1.176819920539856, 1.1444579362869263, 1.135002613067627, 1.1021599769592285, 1.1369946002960205, 1.1269404888153076, 1.0954174995422363, 1.1222219467163086, 1.0581010580062866, 1.145021915435791, 1.071311116218567, 1.1128933429718018, 1.0669652223587036, 1.0565876960754395, 1.1146702766418457, 1.0953928232192993, 1.1271753311157227, 1.0932317972183228, 1.0661178827285767, 1.0444798469543457, 1.0204672813415527, 1.08795964717865, 1.0814549922943115, 1.0728992223739624, 1.0623042583465576, 1.0923744440078735, 1.0658148527145386, 1.031435251235962, 1.0631444454193115, 1.1066699028015137, 1.0429185628890991, 1.0590252876281738, 1.073071002960205, 1.0549137592315674, 1.0577346086502075, 1.041921854019165, 1.062059998512268, 1.044806718826294, 1.0507339239120483, 1.0277938842773438, 1.0332401990890503, 1.0249691009521484, 1.097260594367981, 1.0531126260757446, 1.0454561710357666, 1.052244782447815, 1.0415314435958862, 1.0388447046279907, 1.0653448104858398, 1.0280810594558716, 1.0070394277572632, 1.0019690990447998, 1.0356624126434326, 1.0148468017578125, 0.9851112365722656, 1.0560839176177979, 1.0632798671722412, 1.0424830913543701, 1.0090326070785522, 1.0079481601715088, 1.020960807800293, 1.0015606880187988, 1.0298020839691162, 1.033258080482483, 0.9769415259361267, 1.005272388458252, 0.9893047213554382, 0.9702752828598022, 1.009759783744812, 1.0217289924621582, 1.0367376804351807, 1.0449877977371216, 1.0143221616744995, 1.0272270441055298, 0.9545169472694397, 0.9854423999786377, 0.9338521957397461, 0.9725931286811829, 1.031445026397705, 1.0203595161437988, 0.9765403270721436, 0.9886261820793152, 0.977706253528595, 1.0211162567138672, 0.9779174327850342, 0.9723460674285889, 0.9756026864051819, 0.9548981189727783, 0.9844147562980652, 0.9693609476089478, 1.015467882156372, 0.9634175896644592, 0.9599683880805969, 0.9840601682662964, 0.9614827036857605, 0.9539831280708313, 1.0323657989501953, 0.9587183594703674, 0.9905627369880676, 0.9401179552078247, 0.9964290261268616, 0.9694064855575562, 0.9036505818367004, 0.9368289113044739, 0.9898818731307983, 1.003176212310791, 0.9072144031524658, 1.0221718549728394, 0.9936171770095825, 0.9869346618652344, 0.9218440651893616, 0.9431139230728149, 0.9377160668373108, 0.9564927816390991, 0.9846739172935486, 0.9801074862480164, 0.9965552687644958, 0.9262515902519226, 0.9505006074905396, 0.9331085681915283, 0.9686301350593567, 0.9633806943893433, 0.9596667289733887, 0.8819838166236877, 0.8716541528701782, 0.9280248284339905, 0.942160427570343, 0.9629275798797607, 0.9724134206771851, 0.9816722273826599, 0.9828073978424072, 0.9800280332565308, 0.8904901146888733, 0.9021048545837402, 0.9045128226280212, 0.9432451725006104, 0.9328497648239136, 0.9276496767997742, 0.9136160016059875, 0.9396807551383972, 0.9141830801963806, 0.9293264150619507, 0.8895149827003479, 0.927711009979248, 0.9232009649276733, 0.8827953934669495, 0.8973512649536133, 0.9182061553001404, 0.9002174139022827, 0.9204707741737366, 0.913491427898407, 0.9159440398216248, 0.8651007413864136, 0.9241838455200195, 0.8614627718925476, 0.943705141544342, 0.8976435661315918, 0.9227647185325623, 0.8929766416549683, 0.8856300711631775, 0.8784841299057007, 0.8684089183807373, 0.9268943071365356, 0.8563792705535889, 0.8969572186470032, 0.903371274471283, 0.8551511764526367, 0.8628023266792297, 0.8861611485481262, 0.9088961482048035, 0.8872848153114319, 0.8948448896408081, 0.8846031427383423, 0.8761460781097412, 0.8351300358772278, 0.864956796169281, 0.8571327924728394, 0.9325722455978394, 0.8621909618377686, 0.8768951296806335, 0.8311862945556641, 0.8845886588096619, 0.8567072749137878, 0.8712321519851685, 0.8401355743408203, 0.782669723033905, 0.8611539602279663, 0.8527603149414062, 0.8519423007965088, 0.8754191994667053, 0.8376612663269043, 0.8525954484939575, 0.8512539267539978, 0.8458977937698364, 0.8514487147331238, 0.8847209811210632, 0.8207553029060364, 0.8673360347747803, 0.8420796990394592, 0.8465592265129089, 0.826216995716095, 0.8353025317192078, 0.8347958326339722, 0.8193562626838684, 0.8630731701850891, 0.8171973824501038, 0.8254441618919373, 0.8173123002052307, 0.8628208637237549, 0.8071770071983337, 0.8684903979301453, 0.8509913086891174, 0.8041772842407227, 0.8545263409614563, 0.8106364607810974, 0.8023131489753723, 0.8728472590446472, 0.7952834367752075, 0.7818722724914551, 0.7873514294624329, 0.7735207676887512, 0.8338102102279663, 0.800463855266571, 0.7986652851104736, 0.8030408024787903, 0.7469226121902466, 0.814940333366394, 0.8142125010490417, 0.7923153638839722, 0.7377878427505493, 0.8232533931732178, 0.7839972972869873, 0.809356153011322, 0.7973650693893433, 0.773845911026001, 0.8025988936424255, 0.8142495155334473, 0.7777587175369263, 0.7851992845535278, 0.7498127818107605, 0.8048704862594604, 0.8059808015823364, 0.7589330077171326, 0.7635746002197266, 0.7980225682258606, 0.8389768600463867, 0.7669711112976074, 0.7711759805679321, 0.8189694285392761, 0.7590320706367493, 0.7497916221618652, 0.7728052735328674, 0.7132178544998169, 0.8004099726676941, 0.7608174085617065, 0.8005462288856506, 0.7580879330635071, 0.7705683708190918, 0.7251433730125427, 0.7388266324996948, 0.7716055512428284, 0.7645838260650635, 0.7790837287902832, 0.7722604870796204, 0.7692971229553223, 0.7778341174125671, 0.7655901312828064, 0.7592247128486633, 0.740440845489502, 0.7448546290397644, 0.7725114226341248, 0.7595998644828796, 0.7403860688209534, 0.7571578025817871, 0.7463160753250122, 0.7496543526649475, 0.7363705039024353, 0.7079067230224609, 0.7302257418632507, 0.7513023614883423, 0.7194055914878845, 0.7679688334465027, 0.706447184085846, 0.7573496103286743, 0.7673699259757996, 0.7144200801849365, 0.7670286297798157, 0.725495457649231, 0.6810833811759949, 0.7456190586090088, 0.7478892803192139, 0.7343894839286804, 0.7232798337936401, 0.7386061549186707, 0.7336935997009277, 0.7376851439476013, 0.7398577332496643, 0.7386005520820618, 0.7532640099525452, 0.7265220880508423, 0.7393127083778381, 0.7568772435188293, 0.7339179515838623, 0.7495379447937012, 0.7329300045967102, 0.7046588063240051, 0.6983739733695984, 0.7520650029182434, 0.6962326169013977, 0.7462846636772156, 0.7041400074958801, 0.6952909231185913, 0.7125009894371033, 0.7086904048919678, 0.7119784355163574, 0.7261951565742493, 0.6876853108406067, 0.7017242312431335, 0.7435211539268494, 0.7113695740699768, 0.7606698870658875, 0.7115224003791809, 0.7029397487640381, 0.7136968970298767, 0.7222996354103088, 0.7245602011680603, 0.7144078612327576, 0.7755715250968933, 0.669011116027832, 0.7506664395332336, 0.6952720880508423, 0.7118050456047058, 0.7044669389724731, 0.6868329644203186, 0.7212191820144653, 0.7164889574050903, 0.673394501209259, 0.7157304883003235, 0.7172321081161499, 0.7006426453590393, 0.6967864632606506, 0.7071636319160461, 0.6884975433349609, 0.6930884718894958, 0.7027508616447449, 0.689065158367157, 0.6685864329338074, 0.7079479098320007, 0.7174201607704163, 0.6980125904083252, 0.6881361603736877, 0.7055683732032776, 0.6766249537467957, 0.7017287611961365, 0.6751925349235535, 0.6819112300872803, 0.7593637108802795, 0.6772128343582153, 0.7280768752098083, 0.6768978834152222, 0.7068824172019958]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vmoNb8RWuo94"
      },
      "source": [
        "## Exercise 7"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MrKoM4zzurhH",
        "outputId": "a7130c1a-565d-49e5-add5-e0fec3c16d44"
      },
      "source": [
        "average_mae_history = [np.mean([x[i] for x in all_mae_histories]) for i in range(num_epochs)]\n",
        "average_mae_history_train = [np.mean([x[i] for x in all_mae_history_train]) for i in range(num_epochs)]\n",
        "\n",
        "average_mae_history"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[4.403945088386536,\n",
              " 3.3415854573249817,\n",
              " 3.1253063082695007,\n",
              " 2.854921281337738,\n",
              " 2.818695068359375,\n",
              " 2.5924323201179504,\n",
              " 2.673301577568054,\n",
              " 2.5962324142456055,\n",
              " 2.6547186970710754,\n",
              " 2.7203423380851746,\n",
              " 2.5221301317214966,\n",
              " 2.5474504828453064,\n",
              " 2.552985191345215,\n",
              " 2.687986373901367,\n",
              " 2.485670268535614,\n",
              " 2.5446648001670837,\n",
              " 2.5103413462638855,\n",
              " 2.471744656562805,\n",
              " 2.5236250460147858,\n",
              " 2.5500981211662292,\n",
              " 2.5373569428920746,\n",
              " 2.467947781085968,\n",
              " 2.5091936588287354,\n",
              " 2.4590787291526794,\n",
              " 2.382663279771805,\n",
              " 2.441138416528702,\n",
              " 2.4973554611206055,\n",
              " 2.58024263381958,\n",
              " 2.4058365523815155,\n",
              " 2.40878364443779,\n",
              " 2.375685840845108,\n",
              " 2.393559992313385,\n",
              " 2.4490518867969513,\n",
              " 2.3677046298980713,\n",
              " 2.4302245378494263,\n",
              " 2.384052574634552,\n",
              " 2.406167060136795,\n",
              " 2.4338082671165466,\n",
              " 2.4118300676345825,\n",
              " 2.4367571473121643,\n",
              " 2.43359375,\n",
              " 2.4127676486968994,\n",
              " 2.4933879673480988,\n",
              " 2.5319701433181763,\n",
              " 2.391647845506668,\n",
              " 2.361752539873123,\n",
              " 2.394212305545807,\n",
              " 2.4018789529800415,\n",
              " 2.44615775346756,\n",
              " 2.330051004886627,\n",
              " 2.382799804210663,\n",
              " 2.414953112602234,\n",
              " 2.461215376853943,\n",
              " 2.4048754572868347,\n",
              " 2.3370778560638428,\n",
              " 2.410009503364563,\n",
              " 2.368305116891861,\n",
              " 2.7009560465812683,\n",
              " 2.4105020463466644,\n",
              " 2.4721453189849854,\n",
              " 2.3666819632053375,\n",
              " 2.41877818107605,\n",
              " 2.3737730383872986,\n",
              " 2.455905020236969,\n",
              " 2.594913065433502,\n",
              " 2.437529444694519,\n",
              " 2.5341766476631165,\n",
              " 2.4340329468250275,\n",
              " 2.5962212681770325,\n",
              " 2.5635419487953186,\n",
              " 2.4731645584106445,\n",
              " 2.343983620405197,\n",
              " 2.484933614730835,\n",
              " 2.440971314907074,\n",
              " 2.456498682498932,\n",
              " 2.4286330342292786,\n",
              " 2.4303810596466064,\n",
              " 2.510317027568817,\n",
              " 2.5157249569892883,\n",
              " 2.4437251687049866,\n",
              " 2.455302983522415,\n",
              " 2.479544520378113,\n",
              " 2.6246452927589417,\n",
              " 2.574982702732086,\n",
              " 2.5850566625595093,\n",
              " 2.499554395675659,\n",
              " 2.4845540523529053,\n",
              " 2.4722540378570557,\n",
              " 2.4350696206092834,\n",
              " 2.522601544857025,\n",
              " 2.476362109184265,\n",
              " 2.498486638069153,\n",
              " 2.5004770755767822,\n",
              " 2.551355481147766,\n",
              " 2.529892146587372,\n",
              " 2.582621395587921,\n",
              " 2.50309956073761,\n",
              " 2.493318021297455,\n",
              " 2.452640950679779,\n",
              " 2.552828311920166,\n",
              " 2.5032700300216675,\n",
              " 2.6153488755226135,\n",
              " 2.6078267097473145,\n",
              " 2.8001474142074585,\n",
              " 2.485382378101349,\n",
              " 2.476457178592682,\n",
              " 2.549259066581726,\n",
              " 2.497214138507843,\n",
              " 2.534441113471985,\n",
              " 2.5002992153167725,\n",
              " 2.4835455417633057,\n",
              " 2.393647849559784,\n",
              " 2.5503889322280884,\n",
              " 2.570798397064209,\n",
              " 2.5081213116645813,\n",
              " 2.690523326396942,\n",
              " 2.6311643719673157,\n",
              " 2.524570405483246,\n",
              " 2.854516565799713,\n",
              " 2.5440755486488342,\n",
              " 2.571018695831299,\n",
              " 2.553138852119446,\n",
              " 2.6783854365348816,\n",
              " 2.5586068630218506,\n",
              " 2.5556430220603943,\n",
              " 2.723463714122772,\n",
              " 2.5115031003952026,\n",
              " 2.6304147839546204,\n",
              " 2.7232993841171265,\n",
              " 2.5333439111709595,\n",
              " 2.6036872267723083,\n",
              " 2.632560968399048,\n",
              " 2.61737322807312,\n",
              " 2.5835151076316833,\n",
              " 2.649708867073059,\n",
              " 2.589756727218628,\n",
              " 2.606197953224182,\n",
              " 2.724053144454956,\n",
              " 2.7035622000694275,\n",
              " 2.668363034725189,\n",
              " 2.5757845640182495,\n",
              " 2.602528750896454,\n",
              " 2.6754924654960632,\n",
              " 2.6075605750083923,\n",
              " 2.6439459919929504,\n",
              " 2.578913688659668,\n",
              " 2.6887614727020264,\n",
              " 2.774149715900421,\n",
              " 2.702347159385681,\n",
              " 2.738828480243683,\n",
              " 2.6277642250061035,\n",
              " 2.6278868913650513,\n",
              " 2.7019302248954773,\n",
              " 2.6015522480010986,\n",
              " 2.7779428958892822,\n",
              " 2.573288917541504,\n",
              " 2.6521936655044556,\n",
              " 2.7720780968666077,\n",
              " 2.681950330734253,\n",
              " 2.739397168159485,\n",
              " 2.646582782268524,\n",
              " 2.6752588152885437,\n",
              " 2.6425973176956177,\n",
              " 2.692572236061096,\n",
              " 2.590094029903412,\n",
              " 2.6325382590293884,\n",
              " 2.543538272380829,\n",
              " 2.6703569889068604,\n",
              " 2.7769786715507507,\n",
              " 2.6422321796417236,\n",
              " 2.6299972534179688,\n",
              " 2.592662513256073,\n",
              " 2.7306382060050964,\n",
              " 2.656027138233185,\n",
              " 2.743320107460022,\n",
              " 2.635738492012024,\n",
              " 2.8104902505874634,\n",
              " 2.798959791660309,\n",
              " 2.7012794613838196,\n",
              " 2.698594570159912,\n",
              " 2.6401607990264893,\n",
              " 2.591535449028015,\n",
              " 2.720624566078186,\n",
              " 2.5754615664482117,\n",
              " 2.6439844369888306,\n",
              " 2.6669012308120728,\n",
              " 2.6411438584327698,\n",
              " 2.743080258369446,\n",
              " 2.6726571321487427,\n",
              " 2.635783851146698,\n",
              " 2.7319833636283875,\n",
              " 2.665128707885742,\n",
              " 2.588948130607605,\n",
              " 2.712338149547577,\n",
              " 2.6093665957450867,\n",
              " 2.7049283385276794,\n",
              " 2.6540207862854004,\n",
              " 2.696760416030884,\n",
              " 2.7301039695739746,\n",
              " 2.704158127307892,\n",
              " 2.803207576274872,\n",
              " 2.7002972960472107,\n",
              " 2.735399842262268,\n",
              " 2.7127789855003357,\n",
              " 2.624213993549347,\n",
              " 2.622656285762787,\n",
              " 2.8006129264831543,\n",
              " 2.6652451753616333,\n",
              " 2.640512228012085,\n",
              " 2.6533990502357483,\n",
              " 2.8032678365707397,\n",
              " 2.738123595714569,\n",
              " 2.7098993062973022,\n",
              " 2.609067678451538,\n",
              " 2.6408957839012146,\n",
              " 2.7732405066490173,\n",
              " 2.688219726085663,\n",
              " 2.763346016407013,\n",
              " 2.687524139881134,\n",
              " 2.738192856311798,\n",
              " 2.6362435817718506,\n",
              " 2.664613366127014,\n",
              " 2.757658064365387,\n",
              " 2.751167058944702,\n",
              " 2.652817964553833,\n",
              " 2.479825496673584,\n",
              " 2.6532015800476074,\n",
              " 2.6851093769073486,\n",
              " 2.6538484692573547,\n",
              " 2.7251698970794678,\n",
              " 2.6561191082000732,\n",
              " 2.607457935810089,\n",
              " 2.59555321931839,\n",
              " 2.7737274765968323,\n",
              " 2.682401657104492,\n",
              " 2.751438021659851,\n",
              " 2.6762553453445435,\n",
              " 2.65444678068161,\n",
              " 2.868718385696411,\n",
              " 2.7269952297210693,\n",
              " 2.6888079047203064,\n",
              " 2.612117052078247,\n",
              " 2.76195228099823,\n",
              " 2.6300508975982666,\n",
              " 2.674535036087036,\n",
              " 2.815243422985077,\n",
              " 2.600246787071228,\n",
              " 2.747299313545227,\n",
              " 2.65428364276886,\n",
              " 2.7120935916900635,\n",
              " 2.8685030341148376,\n",
              " 2.6105664372444153,\n",
              " 2.7546924352645874,\n",
              " 2.778220295906067,\n",
              " 2.670442581176758,\n",
              " 2.758910357952118,\n",
              " 2.694258391857147,\n",
              " 2.665067493915558,\n",
              " 2.6742407083511353,\n",
              " 2.702754855155945,\n",
              " 2.681080400943756,\n",
              " 2.751184046268463,\n",
              " 2.6973268389701843,\n",
              " 2.6022353172302246,\n",
              " 2.688340485095978,\n",
              " 2.6476802825927734,\n",
              " 2.7213847041130066,\n",
              " 2.851162612438202,\n",
              " 2.7171342372894287,\n",
              " 2.739564538002014,\n",
              " 2.6724549531936646,\n",
              " 2.737556755542755,\n",
              " 2.6570006608963013,\n",
              " 2.7822681665420532,\n",
              " 2.754770338535309,\n",
              " 2.790509521961212,\n",
              " 2.77552193403244,\n",
              " 2.747909903526306,\n",
              " 2.6823058128356934,\n",
              " 2.7568503618240356,\n",
              " 2.730987548828125,\n",
              " 2.765026092529297,\n",
              " 2.8357731103897095,\n",
              " 2.7400776743888855,\n",
              " 2.811174690723419,\n",
              " 2.7215147614479065,\n",
              " 2.6591637134552,\n",
              " 2.6625874638557434,\n",
              " 2.745215654373169,\n",
              " 2.7605607509613037,\n",
              " 2.8402270078659058,\n",
              " 2.7680171132087708,\n",
              " 2.87479966878891,\n",
              " 2.74161958694458,\n",
              " 2.690101742744446,\n",
              " 2.7491350173950195,\n",
              " 2.8144631385803223,\n",
              " 2.6976397037506104,\n",
              " 2.665741205215454,\n",
              " 2.7051331400871277,\n",
              " 2.7719132900238037,\n",
              " 2.699254035949707,\n",
              " 2.7792468070983887,\n",
              " 2.7612287998199463,\n",
              " 2.719263195991516,\n",
              " 2.7157418727874756,\n",
              " 2.7898126244544983,\n",
              " 2.758421301841736,\n",
              " 2.7908947467803955,\n",
              " 2.7692959904670715,\n",
              " 2.6908233761787415,\n",
              " 2.7289523482322693,\n",
              " 2.8035256266593933,\n",
              " 2.7432538866996765,\n",
              " 2.7702401876449585,\n",
              " 2.843990921974182,\n",
              " 2.712738573551178,\n",
              " 2.661742925643921,\n",
              " 2.7818350791931152,\n",
              " 2.6996833086013794,\n",
              " 2.771485447883606,\n",
              " 2.734437346458435,\n",
              " 2.785898983478546,\n",
              " 2.853883743286133,\n",
              " 2.765779197216034,\n",
              " 2.697154939174652,\n",
              " 2.8484453558921814,\n",
              " 2.781346023082733,\n",
              " 2.79425448179245,\n",
              " 2.715206742286682,\n",
              " 2.822773039340973,\n",
              " 2.698056161403656,\n",
              " 2.8258641362190247,\n",
              " 2.6619166135787964,\n",
              " 2.810885787010193,\n",
              " 2.754721462726593,\n",
              " 2.8635663986206055,\n",
              " 2.7687122225761414,\n",
              " 2.8742511868476868,\n",
              " 2.857043445110321,\n",
              " 2.9394447803497314,\n",
              " 2.762251913547516,\n",
              " 2.80417001247406,\n",
              " 2.8333892226219177,\n",
              " 2.8446196913719177,\n",
              " 2.723175883293152,\n",
              " 2.866803765296936,\n",
              " 2.7104580402374268,\n",
              " 2.7559197545051575,\n",
              " 2.7396386861801147,\n",
              " 2.785478949546814,\n",
              " 2.7591962814331055,\n",
              " 2.818122386932373,\n",
              " 2.791373372077942,\n",
              " 2.7838088274002075,\n",
              " 2.791046440601349,\n",
              " 2.748839259147644,\n",
              " 2.7616634368896484,\n",
              " 2.745444357395172,\n",
              " 2.7123003602027893,\n",
              " 2.7311280965805054,\n",
              " 2.74199241399765,\n",
              " 2.774479389190674,\n",
              " 2.7168803811073303,\n",
              " 2.7675596475601196,\n",
              " 2.863363027572632,\n",
              " 2.8218915462493896,\n",
              " 2.6755102276802063,\n",
              " 2.768060624599457,\n",
              " 2.753804862499237,\n",
              " 2.842461347579956,\n",
              " 2.7494643926620483,\n",
              " 2.8323134779930115,\n",
              " 2.7484946250915527,\n",
              " 2.737110137939453,\n",
              " 2.698050379753113,\n",
              " 2.7179861068725586,\n",
              " 2.7059725522994995,\n",
              " 2.804746448993683,\n",
              " 2.8374491930007935,\n",
              " 2.697709858417511,\n",
              " 2.692439615726471,\n",
              " 2.7096264958381653,\n",
              " 2.738462805747986,\n",
              " 2.7333009243011475,\n",
              " 2.7809808254241943,\n",
              " 2.7124173045158386,\n",
              " 2.686805248260498,\n",
              " 2.7656160593032837,\n",
              " 2.8829535841941833,\n",
              " 2.712611734867096,\n",
              " 2.765444755554199,\n",
              " 2.751456320285797,\n",
              " 2.729998230934143,\n",
              " 2.7271331548690796,\n",
              " 2.8593334555625916,\n",
              " 2.786090135574341,\n",
              " 2.8318912982940674,\n",
              " 2.727755308151245,\n",
              " 2.756569504737854,\n",
              " 2.715065062046051,\n",
              " 2.749643623828888,\n",
              " 2.7286258339881897,\n",
              " 2.7259796261787415,\n",
              " 2.76028573513031,\n",
              " 2.7335948944091797,\n",
              " 2.7600753903388977,\n",
              " 2.7869070768356323,\n",
              " 2.73389995098114,\n",
              " 2.7623842358589172,\n",
              " 2.866934895515442,\n",
              " 2.8100545406341553,\n",
              " 2.61291366815567,\n",
              " 2.7437759041786194,\n",
              " 2.795789897441864,\n",
              " 2.8391001224517822,\n",
              " 2.7263208627700806,\n",
              " 2.767510712146759,\n",
              " 2.8595845103263855,\n",
              " 2.7566481828689575,\n",
              " 2.7567503452301025,\n",
              " 2.746731221675873,\n",
              " 2.788304090499878,\n",
              " 2.744397819042206,\n",
              " 2.7773711681365967,\n",
              " 2.839707314968109,\n",
              " 2.862252354621887,\n",
              " 2.6961905360221863,\n",
              " 2.9147598147392273,\n",
              " 2.8653514981269836,\n",
              " 2.802698075771332,\n",
              " 2.762350082397461,\n",
              " 2.789806842803955,\n",
              " 2.813297688961029,\n",
              " 2.7391690015792847,\n",
              " 2.6964123845100403,\n",
              " 2.7648454904556274,\n",
              " 2.8523009419441223,\n",
              " 2.7711828351020813,\n",
              " 2.6674538254737854,\n",
              " 2.777558743953705,\n",
              " 2.7662306427955627,\n",
              " 2.7586191296577454,\n",
              " 2.852649748325348,\n",
              " 2.7607256174087524,\n",
              " 2.7209303975105286,\n",
              " 2.7901429533958435,\n",
              " 2.640514850616455,\n",
              " 2.7116507291793823,\n",
              " 2.7294029593467712,\n",
              " 2.716506004333496,\n",
              " 2.800389587879181,\n",
              " 2.8044272661209106,\n",
              " 2.7688429951667786,\n",
              " 2.7686806321144104,\n",
              " 2.692942798137665,\n",
              " 2.702096402645111,\n",
              " 2.6882513761520386,\n",
              " 2.7193073630332947,\n",
              " 2.6651893258094788,\n",
              " 2.788230836391449,\n",
              " 2.756286859512329,\n",
              " 2.789780557155609,\n",
              " 2.830331861972809,\n",
              " 2.71263587474823,\n",
              " 2.796016752719879,\n",
              " 2.8850526213645935,\n",
              " 2.859853744506836,\n",
              " 2.7199390530586243,\n",
              " 2.6967108249664307,\n",
              " 2.787577509880066,\n",
              " 2.818570852279663,\n",
              " 2.733142077922821,\n",
              " 2.812241017818451,\n",
              " 2.7871994376182556,\n",
              " 2.8183229565620422,\n",
              " 2.7270761132240295,\n",
              " 2.745446503162384,\n",
              " 2.753419041633606,\n",
              " 2.6500848531723022,\n",
              " 2.7744483947753906,\n",
              " 2.7693710923194885,\n",
              " 2.7013978958129883,\n",
              " 2.7518084049224854,\n",
              " 2.7394416332244873,\n",
              " 2.8167688250541687,\n",
              " 2.7857413291931152,\n",
              " 2.78383731842041,\n",
              " 2.7365203499794006,\n",
              " 2.7811021208763123,\n",
              " 2.760931074619293,\n",
              " 2.7117100954055786,\n",
              " 2.7418338656425476,\n",
              " 2.7480239868164062,\n",
              " 2.805993139743805,\n",
              " 2.685314416885376,\n",
              " 2.6619889736175537,\n",
              " 2.7824397683143616,\n",
              " 2.7477598190307617,\n",
              " 2.6902060508728027]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 17
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U-LhNB_Gu-3w"
      },
      "source": [
        "## Exercise 8"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "CR248AMHvAPO",
        "outputId": "add58022-c5e8-4cd1-846d-62dd3cfc980a"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.plot(range(10, len(average_mae_history) + 1), average_mae_history[9:], label ='val')\n",
        "plt.plot(range(10, len(average_mae_history) + 1), average_mae_history_train[9:], label ='train')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Validation MAE')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "liDFOhIcvaay"
      },
      "source": [
        "## Exercise 9"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "r_LJAnzOvbnu",
        "outputId": "36be6e3d-c6c8-4710-e395-bf29c1814293"
      },
      "source": [
        "def smooth_curve(points, factor=0.9):\n",
        "  smoothed_points = []\n",
        "  for point in points:\n",
        "    if smoothed_points:\n",
        "      previous = smoothed_points[-1]\n",
        "      smoothed_points.append(previous * factor + point * (1 - factor))\n",
        "    else:\n",
        "      smoothed_points.append(point)\n",
        "  return smoothed_points\n",
        "\n",
        "smooth_mae_history = smooth_curve(average_mae_history[10:])\n",
        "smooth_mae_history_train = smooth_curve(average_mae_history_train[10:])\n",
        "\n",
        "plt.plot(range(1, len(smooth_mae_history) + 1), smooth_mae_history,label ='val')\n",
        "plt.plot(range(1, len(smooth_mae_history_train) + 1), smooth_mae_history_train, label ='train')\n",
        "\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Validation MAE')\n",
        "plt.legend()\n",
        "plt.show()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}