Deep Learning/Introduction to Neural Networks with Tensorflow/Regression_Classification_using_Sequential_API.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Regression/Classification using Sequential API.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "#Classification MLP using the **Sequential API**"
      ],
      "metadata": {
        "id": "jnHZVmTqOkU_"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "zjboHoY8qTDk"
      },
      "outputs": [],
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow.keras import datasets"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "###Download Data"
      ],
      "metadata": {
        "id": "G4mx2FG9qt5f"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "fashion_mnist = datasets.fashion_mnist\n",
        "(x_train_full, y_train_full), (x_test, y_test) = fashion_mnist.load_data()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OluqQrUlqaJa",
        "outputId": "12448c18-cd9d-48f7-e9a2-1de9f3a0d9b3"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n",
            "32768/29515 [=================================] - 0s 0us/step\n",
            "40960/29515 [=========================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz\n",
            "26427392/26421880 [==============================] - 0s 0us/step\n",
            "26435584/26421880 [==============================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n",
            "16384/5148 [===============================================================================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz\n",
            "4423680/4422102 [==============================] - 0s 0us/step\n",
            "4431872/4422102 [==============================] - 0s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x_train_full.shape, x_test.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9BJ4J1ejqiLq",
        "outputId": "daa701cd-c611-4d18-ed96-bb688d7998bf"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((60000, 28, 28), (10000, 28, 28))"
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "The difference of dowloading the dataset using tensorflow instead of scikit learn is that the image are rapresented as a 28x28 matric instead of a 1D vector, and the pixel intensities are integer instead of floats."
      ],
      "metadata": {
        "id": "UySzJZ1eq7YH"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "###Create a validation set\n",
        "also rescale the features and convert to floats"
      ],
      "metadata": {
        "id": "8A-gakLPrQ4h"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x_val, x_train = x_train_full[:5_000]/255., x_train_full[5_000:]/255. \n",
        "y_val, y_train = y_train_full[:5_000], y_train_full[5_000:]"
      ],
      "metadata": {
        "id": "ev27G4Xiq2FU"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "We need the list of classes to know what we are dealing with"
      ],
      "metadata": {
        "id": "3PWQLLB2sAm-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class_names = [\"T-shirt\",\"Trousers\",\"Pullover\",\"Dress\",\"Coat\",\"Sandal\",\"Shirt\",\"Sneaker\",\"Bg\",\"Ankle boot\"]"
      ],
      "metadata": {
        "id": "yBxQRYz3rl5O"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt"
      ],
      "metadata": {
        "id": "HHVUihAHsaqL"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plt.imshow(x_train[0], cmap = \"gray\")\n",
        "class_names[y_train[0]]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "CuZdM4SPskuS",
        "outputId": "4bf2da20-3c33-40ed-952d-bd530fb39bb8"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'Coat'"
            ]
          },
          "metadata": {},
          "execution_count": 7
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Create the model using Sequential API"
      ],
      "metadata": {
        "id": "nMUuT736suov"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow.keras import Sequential\n",
        "from tensorflow.keras.layers import Dense, Flatten,InputLayer"
      ],
      "metadata": {
        "id": "bIjvRN2EsojS"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = Sequential([\n",
        "  Flatten(input_shape = [28,28]),\n",
        "  Dense(units=300, activation=\"relu\", name = \"dense_1\"),\n",
        "  Dense(units=100, activation=\"relu\"),\n",
        "  Dense(units=10, activation=\"softmax\"),\n",
        "])"
      ],
      "metadata": {
        "id": "xjJI0RL4s5rk"
      },
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Flatten is used as a sort of \"preprocessing\" step to convert 2D data into 1D.\n",
        "Softmax let us interpret the output as a probability distribution over the 10 classes."
      ],
      "metadata": {
        "id": "U3wJO8fwOAQr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OuwM3h1LtEzn",
        "outputId": "24bbb1f2-2763-4db9-a43a-70481b6da96d"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " flatten_1 (Flatten)         (None, 784)               0         \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 300)               235500    \n",
            "                                                                 \n",
            " dense_2 (Dense)             (None, 100)               30100     \n",
            "                                                                 \n",
            " dense_3 (Dense)             (None, 10)                1010      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 266,610\n",
            "Trainable params: 266,610\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Access layers"
      ],
      "metadata": {
        "id": "p0ZKqQDWuKz6"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model.layers"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nG6VnzTNuCaK",
        "outputId": "f518a84a-c7b4-4450-cdc1-ec726d015563"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<keras.layers.core.flatten.Flatten at 0x7f0ed800c450>,\n",
              " <keras.layers.core.dense.Dense at 0x7f0ed800c5d0>,\n",
              " <keras.layers.core.dense.Dense at 0x7f0edb9abe90>,\n",
              " <keras.layers.core.dense.Dense at 0x7f0edc641d10>]"
            ]
          },
          "metadata": {},
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model.get_layer(\"dense_1\").name"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "id": "8fQTI9HxuNtQ",
        "outputId": "0a23a18b-8f9d-4a9b-fb53-9df38e61896f"
      },
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'dense_1'"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Get parameters from a layer"
      ],
      "metadata": {
        "id": "YJ_7fNLwujLm"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "weights, bias = model.layers[1].get_weights()\n",
        "\n",
        "weights.shape, bias.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YL6KgOl2uS-i",
        "outputId": "da5b2b2e-1cab-4a03-9516-8211070e7363"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((784, 300), (300,))"
            ]
          },
          "metadata": {},
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Compiling the model"
      ],
      "metadata": {
        "id": "tArw59zNus9R"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model.compile(optimizer = \"sgd\", loss = \"sparse_categorical_crossentropy\", metrics = [\"accuracy\"])"
      ],
      "metadata": {
        "id": "5erbe98PugBM"
      },
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "We can use categorical_crossentropy if we use one-hot target ecnoding"
      ],
      "metadata": {
        "id": "rKZUW9bSu8gP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow import keras "
      ],
      "metadata": {
        "id": "m9gvU8lEvMzf"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "y_test = keras.utils.to_categorical(y_test)\n",
        "y_val = keras.utils.to_categorical(y_val)\n",
        "y_train = keras.utils.to_categorical(y_train) "
      ],
      "metadata": {
        "id": "H0OBJg8Zu5d2"
      },
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.compile(optimizer = \"sgd\", loss = \"categorical_crossentropy\", metrics = [\"accuracy\"])"
      ],
      "metadata": {
        "id": "h4Co2UlyvQsF"
      },
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(x_train,y_train, validation_data= (x_val,y_val), epochs=10)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Qrq4g16MvgKn",
        "outputId": "9298240c-6010-4087-b3f0-d1a2852d41ef"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "1719/1719 [==============================] - 7s 4ms/step - loss: 0.7164 - accuracy: 0.7702 - val_loss: 0.5430 - val_accuracy: 0.8058\n",
            "Epoch 2/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.4881 - accuracy: 0.8304 - val_loss: 0.4433 - val_accuracy: 0.8542\n",
            "Epoch 3/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.4457 - accuracy: 0.8438 - val_loss: 0.4229 - val_accuracy: 0.8550\n",
            "Epoch 4/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.4185 - accuracy: 0.8551 - val_loss: 0.4043 - val_accuracy: 0.8620\n",
            "Epoch 5/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.3975 - accuracy: 0.8591 - val_loss: 0.3885 - val_accuracy: 0.8608\n",
            "Epoch 6/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.3821 - accuracy: 0.8653 - val_loss: 0.3778 - val_accuracy: 0.8686\n",
            "Epoch 7/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.3680 - accuracy: 0.8700 - val_loss: 0.3755 - val_accuracy: 0.8678\n",
            "Epoch 8/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.3554 - accuracy: 0.8743 - val_loss: 0.3876 - val_accuracy: 0.8596\n",
            "Epoch 9/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.3459 - accuracy: 0.8771 - val_loss: 0.3492 - val_accuracy: 0.8760\n",
            "Epoch 10/10\n",
            "1719/1719 [==============================] - 6s 3ms/step - loss: 0.3354 - accuracy: 0.8795 - val_loss: 0.3461 - val_accuracy: 0.8754\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Let's plot learning curves"
      ],
      "metadata": {
        "id": "fZhnPdqRwB3U"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import pandas as pd"
      ],
      "metadata": {
        "id": "U98wB9bgvpFt"
      },
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "pd.DataFrame(history.history).plot(figsize= (8,5))\n",
        "plt.grid(True)\n",
        "plt.gca().set_ylim(0,1)\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 324
        },
        "id": "fw6hMtYmvuvK",
        "outputId": "acff71b2-0250-4019-ac2f-8c9831a9d233"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Make predictions"
      ],
      "metadata": {
        "id": "x57urzAUwiVy"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np"
      ],
      "metadata": {
        "id": "z2De75hIN6qw"
      },
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_new = x_test[:3]\n",
        "y_new = y_test[:3]\n",
        "y_new = [np.argmax(instance) for instance in y_new]"
      ],
      "metadata": {
        "id": "NUZwgHQfv_0Q"
      },
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "y_pred = model.predict(x_new)\n",
        "y_pred = [np.argmax(instance) for instance in y_pred]"
      ],
      "metadata": {
        "id": "JrZiW5ROwqAN"
      },
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Check if predictions of new instances are equal to the correct labels."
      ],
      "metadata": {
        "id": "D8dV5I8FNLzX"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "[class_names[i] for i in y_pred]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Xj35p_uywswu",
        "outputId": "716cc80b-b527-42c5-8091-2d47655a1d32"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['Ankle boot', 'Pullover', 'Trousers']"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "[class_names[i] for i in y_new]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_AuKKUVIw7ol",
        "outputId": "4bf8ee39-c38a-49e6-a7cb-602fddaa09d0"
      },
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['Ankle boot', 'Pullover', 'Trousers']"
            ]
          },
          "metadata": {},
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Regression MLP using the **Sequential API**"
      ],
      "metadata": {
        "id": "B5QvIPRCOqph"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "We are going to use the california housing dataset from scikit learn, since it is simple to deal with, in fact it only contains numerical features and no missing values."
      ],
      "metadata": {
        "id": "Ts1M8R2ROtDW"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.datasets import fetch_california_housing\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.preprocessing import StandardScaler"
      ],
      "metadata": {
        "id": "T2KuP4P7xJIO"
      },
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "###Download the dataset"
      ],
      "metadata": {
        "id": "FnMw6c7yPLR9"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "california_housing = fetch_california_housing()\n",
        "scaler = StandardScaler()"
      ],
      "metadata": {
        "id": "p5dNV_62O-_2"
      },
      "execution_count": 38,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_train_full, x_test, y_train_full, y_test = train_test_split(california_housing[\"data\"],california_housing[\"target\"])"
      ],
      "metadata": {
        "id": "nZHBwiSHPIuO"
      },
      "execution_count": 48,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "  Let's create a validation set"
      ],
      "metadata": {
        "id": "auhZb24TPtCH"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x_train, x_val, y_train, y_val = train_test_split(x_train_full,y_train_full)"
      ],
      "metadata": {
        "id": "N4iLhGOFPiua"
      },
      "execution_count": 49,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "###Standardization"
      ],
      "metadata": {
        "id": "NYjBXSLTQlrc"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x_train = scaler.fit_transform(x_train)\n",
        "x_val = scaler.transform(x_val)\n",
        "x_test = scaler.transform(x_test)"
      ],
      "metadata": {
        "id": "6rSsgGRhPjTy"
      },
      "execution_count": 50,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Since we want to predict a single real value, we should have a single neuron in output with no activation fucntion, so it is allowed to output any number without costraints.\n",
        "We use less neurons and layers than before because the dataset is quite noisy, and we risk to overfitt"
      ],
      "metadata": {
        "id": "-79VN2LHQ0UL"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model = Sequential([\n",
        "  Dense(units = 30, activation= \"relu\", input_shape = x_train.shape[1:]),\n",
        "  Dense(1)\n",
        "])"
      ],
      "metadata": {
        "id": "bBENu9brQDGm"
      },
      "execution_count": 73,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.compile(loss = \"mean_squared_error\", optimizer = \"sgd\")"
      ],
      "metadata": {
        "id": "QoLv-RF8RlD6"
      },
      "execution_count": 74,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(x_train,y_train, validation_data = (x_val,y_val),epochs = 10)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yuheGuUkRrXT",
        "outputId": "ce506520-1926-4c01-b5c1-8e1bbd81f389"
      },
      "execution_count": 82,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.4050 - val_loss: 0.3928\n",
            "Epoch 2/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.4066 - val_loss: 0.3938\n",
            "Epoch 3/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.3986 - val_loss: 0.3906\n",
            "Epoch 4/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.3949 - val_loss: 0.3817\n",
            "Epoch 5/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.3923 - val_loss: 0.3805\n",
            "Epoch 6/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.5197 - val_loss: 0.5037\n",
            "Epoch 7/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.5212 - val_loss: 0.4211\n",
            "Epoch 8/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.4148 - val_loss: 0.4033\n",
            "Epoch 9/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.4015 - val_loss: 0.3919\n",
            "Epoch 10/10\n",
            "363/363 [==============================] - 1s 2ms/step - loss: 0.3951 - val_loss: 0.3906\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df = pd.DataFrame(history.history)\n",
        "df.plot(figsize = (8,5))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        },
        "id": "Md6jTMi7SOPh",
        "outputId": "034bb232-c589-41c4-f008-b5d9b4f25008"
      },
      "execution_count": 77,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f0ed5635310>"
            ]
          },
          "metadata": {},
          "execution_count": 77
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##Evaluation on test set"
      ],
      "metadata": {
        "id": "yk_ZXe_wScj8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "mse_test = model.evaluate(x_test,y_test)\n",
        "mse_test"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-0ApfO-FRJAm",
        "outputId": "7b2ff7a5-ebc4-4cd4-a54b-149c05fdd66c"
      },
      "execution_count": 83,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "162/162 [==============================] - 0s 1ms/step - loss: 0.3876\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.3876430094242096"
            ]
          },
          "metadata": {},
          "execution_count": 83
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x_new = x_test[:3]\n",
        "model.predict(x_new)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "V4LhcHnmRJTZ",
        "outputId": "734189af-34f7-4b96-8e1e-a4d9afbe4f60"
      },
      "execution_count": 84,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[0.68018824],\n",
              "       [1.7720973 ],\n",
              "       [1.3286722 ]], dtype=float32)"
            ]
          },
          "metadata": {},
          "execution_count": 84
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "y_new = y_test[:3]\n",
        "y_new"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9e0Xq_5ZSx96",
        "outputId": "1b84f9d2-8a6f-4a6e-c338-e2225bbb9d87"
      },
      "execution_count": 85,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0.735, 2.188, 1.411])"
            ]
          },
          "metadata": {},
          "execution_count": 85
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "YmIn1w0YS2si"
      },
      "execution_count": 85,
      "outputs": []
    }
  ]
}