colabs/unit_3-machine_learning_and_analytical_models/Week_15-1-Multinomial_Classification-ANN-spotify_data.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "jYysdyb-CaWM"
   },
   "source": [
    "# Classify spotify artists using an Artificial Neural Network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "FbVhjPpzn6BM"
   },
   "source": [
    "This guide trains a neural network model to classify spotify artists, like taylor swift and beyonce. It is adapted from [this colab](https://www.tensorflow.org/tutorials/keras/classification)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "executionInfo": {
     "elapsed": 3990,
     "status": "ok",
     "timestamp": 1618857320067,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "dzLKpmZICaWN"
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "\n",
    "# Helper libraries\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "yR0EdgrLCaWR"
   },
   "source": [
    "## Import the Spotify Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 28276,
     "status": "ok",
     "timestamp": 1618857344363,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "pdun9pCucVvR",
    "outputId": "6325e061-622e-4975-9f05-0beba69ebf35"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mounted at /content/drive\n"
     ]
    }
   ],
   "source": [
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "executionInfo": {
     "elapsed": 4939,
     "status": "ok",
     "timestamp": 1618857361907,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "7MqDQO0KCaWS"
   },
   "outputs": [],
   "source": [
    "df_taylor  = pd.read_csv(\"/content/drive/MyDrive/python_scratch/spotify_data/taylor.csv\")\n",
    "df_beyonce = pd.read_csv(\"/content/drive/MyDrive/python_scratch/spotify_data/beyonce.csv\")\n",
    "df_beatles = pd.read_csv(\"/content/drive/MyDrive/python_scratch/spotify_data/beatles.csv\")\n",
    "df_nirvana = pd.read_csv(\"/content/drive/MyDrive/python_scratch/spotify_data/nirvana.csv\")\n",
    "df_rolling = pd.read_csv(\"/content/drive/MyDrive/python_scratch/spotify_data/rolling_stones.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "executionInfo": {
     "elapsed": 1084,
     "status": "ok",
     "timestamp": 1618857392395,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "9D100os8uihn"
   },
   "outputs": [],
   "source": [
    "label_names = [  \"Taylor Swift\", \"Beyonce\", \"The Beatles\", \"Nirvana\",  \"The Rolling Stones\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 411
    },
    "executionInfo": {
     "elapsed": 1568,
     "status": "ok",
     "timestamp": 1618857406815,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "T3HTT23Zwhu8",
    "outputId": "154ec21b-0a37-4c88-af75-3f4cc5fff557"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>artist_name</th>\n",
       "      <th>artist_id</th>\n",
       "      <th>album_id</th>\n",
       "      <th>album_type</th>\n",
       "      <th>album_images</th>\n",
       "      <th>album_release_date</th>\n",
       "      <th>album_release_year</th>\n",
       "      <th>album_release_date_precision</th>\n",
       "      <th>danceability</th>\n",
       "      <th>energy</th>\n",
       "      <th>key</th>\n",
       "      <th>loudness</th>\n",
       "      <th>mode</th>\n",
       "      <th>speechiness</th>\n",
       "      <th>acousticness</th>\n",
       "      <th>instrumentalness</th>\n",
       "      <th>liveness</th>\n",
       "      <th>valence</th>\n",
       "      <th>tempo</th>\n",
       "      <th>track_id</th>\n",
       "      <th>analysis_url</th>\n",
       "      <th>time_signature</th>\n",
       "      <th>artists</th>\n",
       "      <th>available_markets</th>\n",
       "      <th>disc_number</th>\n",
       "      <th>duration_ms</th>\n",
       "      <th>explicit</th>\n",
       "      <th>track_href</th>\n",
       "      <th>is_local</th>\n",
       "      <th>track_name</th>\n",
       "      <th>track_preview_url</th>\n",
       "      <th>track_number</th>\n",
       "      <th>type</th>\n",
       "      <th>track_uri</th>\n",
       "      <th>external_urls.spotify</th>\n",
       "      <th>album_name</th>\n",
       "      <th>key_name</th>\n",
       "      <th>mode_name</th>\n",
       "      <th>key_mode</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>Taylor Swift</td>\n",
       "      <td>06HL4z0CvFAxyc27GXpf02</td>\n",
       "      <td>6DEjYFkNZh67HP7R9PSZvv</td>\n",
       "      <td>album</td>\n",
       "      <td>list(height = c(640, 300, 64), url = c(\"https:...</td>\n",
       "      <td>2017-11-10</td>\n",
       "      <td>2017</td>\n",
       "      <td>day</td>\n",
       "      <td>0.613</td>\n",
       "      <td>0.764</td>\n",
       "      <td>2</td>\n",
       "      <td>-6.509</td>\n",
       "      <td>1</td>\n",
       "      <td>0.1360</td>\n",
       "      <td>0.05270</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.1970</td>\n",
       "      <td>0.4170</td>\n",
       "      <td>160.015</td>\n",
       "      <td>2yLa0QULdQr0qAIvVwN6B5</td>\n",
       "      <td>https://api.spotify.com/v1/audio-analysis/2yLa...</td>\n",
       "      <td>4</td>\n",
       "      <td>list(href = \"https://api.spotify.com/v1/artist...</td>\n",
       "      <td>c(\"AD\", \"AE\", \"AR\", \"AT\", \"AU\", \"BE\", \"BG\", \"B...</td>\n",
       "      <td>1</td>\n",
       "      <td>208186</td>\n",
       "      <td>False</td>\n",
       "      <td>https://api.spotify.com/v1/tracks/2yLa0QULdQr0...</td>\n",
       "      <td>False</td>\n",
       "      <td>...Ready For It?</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "      <td>track</td>\n",
       "      <td>spotify:track:2yLa0QULdQr0qAIvVwN6B5</td>\n",
       "      <td>https://open.spotify.com/track/2yLa0QULdQr0qAI...</td>\n",
       "      <td>reputation</td>\n",
       "      <td>D</td>\n",
       "      <td>major</td>\n",
       "      <td>D major</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>Taylor Swift</td>\n",
       "      <td>06HL4z0CvFAxyc27GXpf02</td>\n",
       "      <td>6DEjYFkNZh67HP7R9PSZvv</td>\n",
       "      <td>album</td>\n",
       "      <td>list(height = c(640, 300, 64), url = c(\"https:...</td>\n",
       "      <td>2017-11-10</td>\n",
       "      <td>2017</td>\n",
       "      <td>day</td>\n",
       "      <td>0.649</td>\n",
       "      <td>0.589</td>\n",
       "      <td>2</td>\n",
       "      <td>-6.237</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0558</td>\n",
       "      <td>0.00845</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.1080</td>\n",
       "      <td>0.1510</td>\n",
       "      <td>159.073</td>\n",
       "      <td>2x0WlnmfG39ZuDmstl9xfX</td>\n",
       "      <td>https://api.spotify.com/v1/audio-analysis/2x0W...</td>\n",
       "      <td>4</td>\n",
       "      <td>list(href = c(\"https://api.spotify.com/v1/arti...</td>\n",
       "      <td>c(\"AD\", \"AE\", \"AR\", \"AT\", \"AU\", \"BE\", \"BG\", \"B...</td>\n",
       "      <td>1</td>\n",
       "      <td>244826</td>\n",
       "      <td>False</td>\n",
       "      <td>https://api.spotify.com/v1/tracks/2x0WlnmfG39Z...</td>\n",
       "      <td>False</td>\n",
       "      <td>End Game</td>\n",
       "      <td>NaN</td>\n",
       "      <td>2</td>\n",
       "      <td>track</td>\n",
       "      <td>spotify:track:2x0WlnmfG39ZuDmstl9xfX</td>\n",
       "      <td>https://open.spotify.com/track/2x0WlnmfG39ZuDm...</td>\n",
       "      <td>reputation</td>\n",
       "      <td>D</td>\n",
       "      <td>major</td>\n",
       "      <td>D major</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Taylor Swift</td>\n",
       "      <td>06HL4z0CvFAxyc27GXpf02</td>\n",
       "      <td>6DEjYFkNZh67HP7R9PSZvv</td>\n",
       "      <td>album</td>\n",
       "      <td>list(height = c(640, 300, 64), url = c(\"https:...</td>\n",
       "      <td>2017-11-10</td>\n",
       "      <td>2017</td>\n",
       "      <td>day</td>\n",
       "      <td>0.696</td>\n",
       "      <td>0.602</td>\n",
       "      <td>0</td>\n",
       "      <td>-6.156</td>\n",
       "      <td>0</td>\n",
       "      <td>0.1590</td>\n",
       "      <td>0.06790</td>\n",
       "      <td>0.000021</td>\n",
       "      <td>0.0696</td>\n",
       "      <td>0.3050</td>\n",
       "      <td>82.989</td>\n",
       "      <td>4svZDCRz4cJoneBpjpx8DJ</td>\n",
       "      <td>https://api.spotify.com/v1/audio-analysis/4svZ...</td>\n",
       "      <td>4</td>\n",
       "      <td>list(href = \"https://api.spotify.com/v1/artist...</td>\n",
       "      <td>c(\"AD\", \"AE\", \"AR\", \"AT\", \"AU\", \"BE\", \"BG\", \"B...</td>\n",
       "      <td>1</td>\n",
       "      <td>238253</td>\n",
       "      <td>False</td>\n",
       "      <td>https://api.spotify.com/v1/tracks/4svZDCRz4cJo...</td>\n",
       "      <td>False</td>\n",
       "      <td>I Did Something Bad</td>\n",
       "      <td>NaN</td>\n",
       "      <td>3</td>\n",
       "      <td>track</td>\n",
       "      <td>spotify:track:4svZDCRz4cJoneBpjpx8DJ</td>\n",
       "      <td>https://open.spotify.com/track/4svZDCRz4cJoneB...</td>\n",
       "      <td>reputation</td>\n",
       "      <td>C</td>\n",
       "      <td>minor</td>\n",
       "      <td>C minor</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>Taylor Swift</td>\n",
       "      <td>06HL4z0CvFAxyc27GXpf02</td>\n",
       "      <td>6DEjYFkNZh67HP7R9PSZvv</td>\n",
       "      <td>album</td>\n",
       "      <td>list(height = c(640, 300, 64), url = c(\"https:...</td>\n",
       "      <td>2017-11-10</td>\n",
       "      <td>2017</td>\n",
       "      <td>day</td>\n",
       "      <td>0.615</td>\n",
       "      <td>0.534</td>\n",
       "      <td>9</td>\n",
       "      <td>-6.719</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0386</td>\n",
       "      <td>0.10600</td>\n",
       "      <td>0.000018</td>\n",
       "      <td>0.0607</td>\n",
       "      <td>0.1930</td>\n",
       "      <td>135.917</td>\n",
       "      <td>1R0a2iXumgCiFb7HEZ7gUE</td>\n",
       "      <td>https://api.spotify.com/v1/audio-analysis/1R0a...</td>\n",
       "      <td>4</td>\n",
       "      <td>list(href = \"https://api.spotify.com/v1/artist...</td>\n",
       "      <td>c(\"AD\", \"AE\", \"AR\", \"AT\", \"AU\", \"BE\", \"BG\", \"B...</td>\n",
       "      <td>1</td>\n",
       "      <td>236413</td>\n",
       "      <td>False</td>\n",
       "      <td>https://api.spotify.com/v1/tracks/1R0a2iXumgCi...</td>\n",
       "      <td>False</td>\n",
       "      <td>Don’t Blame Me</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4</td>\n",
       "      <td>track</td>\n",
       "      <td>spotify:track:1R0a2iXumgCiFb7HEZ7gUE</td>\n",
       "      <td>https://open.spotify.com/track/1R0a2iXumgCiFb7...</td>\n",
       "      <td>reputation</td>\n",
       "      <td>A</td>\n",
       "      <td>minor</td>\n",
       "      <td>A minor</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>Taylor Swift</td>\n",
       "      <td>06HL4z0CvFAxyc27GXpf02</td>\n",
       "      <td>6DEjYFkNZh67HP7R9PSZvv</td>\n",
       "      <td>album</td>\n",
       "      <td>list(height = c(640, 300, 64), url = c(\"https:...</td>\n",
       "      <td>2017-11-10</td>\n",
       "      <td>2017</td>\n",
       "      <td>day</td>\n",
       "      <td>0.750</td>\n",
       "      <td>0.404</td>\n",
       "      <td>9</td>\n",
       "      <td>-10.178</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0682</td>\n",
       "      <td>0.21600</td>\n",
       "      <td>0.000357</td>\n",
       "      <td>0.0911</td>\n",
       "      <td>0.0499</td>\n",
       "      <td>95.045</td>\n",
       "      <td>6NFyWDv5CjfwuzoCkw47Xf</td>\n",
       "      <td>https://api.spotify.com/v1/audio-analysis/6NFy...</td>\n",
       "      <td>4</td>\n",
       "      <td>list(href = \"https://api.spotify.com/v1/artist...</td>\n",
       "      <td>c(\"AD\", \"AE\", \"AR\", \"AT\", \"AU\", \"BE\", \"BG\", \"B...</td>\n",
       "      <td>1</td>\n",
       "      <td>232253</td>\n",
       "      <td>False</td>\n",
       "      <td>https://api.spotify.com/v1/tracks/6NFyWDv5Cjfw...</td>\n",
       "      <td>False</td>\n",
       "      <td>Delicate</td>\n",
       "      <td>NaN</td>\n",
       "      <td>5</td>\n",
       "      <td>track</td>\n",
       "      <td>spotify:track:6NFyWDv5CjfwuzoCkw47Xf</td>\n",
       "      <td>https://open.spotify.com/track/6NFyWDv5Cjfwuzo...</td>\n",
       "      <td>reputation</td>\n",
       "      <td>A</td>\n",
       "      <td>minor</td>\n",
       "      <td>A minor</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0   artist_name               artist_id  ... key_name mode_name key_mode\n",
       "0           1  Taylor Swift  06HL4z0CvFAxyc27GXpf02  ...        D     major  D major\n",
       "1           2  Taylor Swift  06HL4z0CvFAxyc27GXpf02  ...        D     major  D major\n",
       "2           3  Taylor Swift  06HL4z0CvFAxyc27GXpf02  ...        C     minor  C minor\n",
       "3           4  Taylor Swift  06HL4z0CvFAxyc27GXpf02  ...        A     minor  A minor\n",
       "4           5  Taylor Swift  06HL4z0CvFAxyc27GXpf02  ...        A     minor  A minor\n",
       "\n",
       "[5 rows x 40 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_taylor.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "executionInfo": {
     "elapsed": 1154,
     "status": "ok",
     "timestamp": 1618857854837,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "PRFaYTS7cf9m"
   },
   "outputs": [],
   "source": [
    "feat_taylor = df_taylor.loc[:, \"danceability\":\"valence\"]\n",
    "feat_beyonce = df_beyonce.loc[:,\"danceability\":\"valence\"]\n",
    "feat_beatles = df_beatles.loc[:,\"danceability\":\"valence\"]\n",
    "feat_nirvana = df_nirvana.loc[:,\"danceability\":\"valence\"]\n",
    "feat_rolling = df_rolling.loc[:,\"danceability\":\"valence\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "executionInfo": {
     "elapsed": 1050,
     "status": "ok",
     "timestamp": 1618857922903,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "TuDTLbUXvYB0"
   },
   "outputs": [],
   "source": [
    "features = np.concatenate([\n",
    "                        feat_taylor,\n",
    "                        feat_beyonce,\n",
    "                        feat_beatles, \n",
    "                        feat_nirvana,\n",
    "                        feat_rolling,\n",
    "                      ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1215,
     "status": "ok",
     "timestamp": 1618860359465,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "v0a6IWWkmDJC",
    "outputId": "27bf8076-f25a-491f-d0bc-4306ab0d7e8e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(547, 438, 139, 213, 743)"
      ]
     },
     "execution_count": 56,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(feat_taylor),len(feat_beyonce),len(feat_beatles),len(feat_nirvana),len(feat_rolling),"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "executionInfo": {
     "elapsed": 1044,
     "status": "ok",
     "timestamp": 1618857927752,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "7t-t13BMyaag"
   },
   "outputs": [],
   "source": [
    "labels = np.concatenate([\n",
    "                    np.zeros(len(feat_taylor)) +0,\n",
    "                    np.zeros(len(feat_beyonce))+1,\n",
    "                    np.zeros(len(feat_beatles))+2,\n",
    "                    np.zeros(len(feat_nirvana))+3, \n",
    "                    np.zeros(len(feat_rolling))+4\n",
    "                      ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1658,
     "status": "ok",
     "timestamp": 1618857932668,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "ZsrDAYwkdZXi",
    "outputId": "8a2b6f66-1bfa-4f88-c85f-46dbdb2c26f9"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((2080, 10), (2080,))"
      ]
     },
     "execution_count": 9,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features.shape, labels.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "executionInfo": {
     "elapsed": 2029,
     "status": "ok",
     "timestamp": 1618857980788,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "6_arQdsezncm"
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "train_features, test_features, train_labels, test_labels = train_test_split(features, labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "RFB6XemAlZUL"
   },
   "source": [
    "## Plot the first 25 elements\n",
    "\n",
    "To see what it looks like"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 589
    },
    "executionInfo": {
     "elapsed": 2351,
     "status": "ok",
     "timestamp": 1618858139103,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "oZTImqg_CaW1",
    "outputId": "bc589160-35d7-47dd-a9ec-618aa0e31f04"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 25 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,10))\n",
    "i=0\n",
    "for feat, label in zip(train_features, train_labels):\n",
    "    plt.subplot(5,5,i+1)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    plt.grid(False)\n",
    "    plt.plot(feat)\n",
    "    plt.xlabel(label_names[int(label)])\n",
    "    i+=1\n",
    "    if i >= 25:\n",
    "        break\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "59veuiEZCaW4"
   },
   "source": [
    "## Build the model\n",
    "\n",
    "Building the neural network requires configuring the layers of the model, then compiling the model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Gxg1XGm0eOBy"
   },
   "source": [
    "### Set up the layers\n",
    "\n",
    "The basic building block of a neural network is the *layer*. Layers extract representations from the data fed into them. Hopefully, these representations are meaningful for the problem at hand.\n",
    "\n",
    "Most of deep learning consists of chaining together simple layers. Most layers, such as `tf.keras.layers.Dense`, have parameters that are learned during training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1019,
     "status": "ok",
     "timestamp": 1618858467796,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "5tLEL-NApInB",
    "outputId": "da0bccb8-0685-4cbf-f73a-0507988465a0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10,)\n"
     ]
    }
   ],
   "source": [
    "input_shape = features[0].shape\n",
    "print(input_shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "executionInfo": {
     "elapsed": 1433,
     "status": "ok",
     "timestamp": 1618861395272,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "9ODch-OFCaW4"
   },
   "outputs": [],
   "source": [
    "model = tf.keras.Sequential(\n",
    "    [\n",
    "    tf.keras.layers.Input(input_shape),\n",
    "    tf.keras.layers.experimental.preprocessing.Normalization(),\n",
    "    tf.keras.layers.Dense(128, activation='relu'),\n",
    "    tf.keras.layers.Dropout(0.3),\n",
    "    tf.keras.layers.Dense(len(label_names))]\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "gut8A_7rCaW6"
   },
   "source": [
    "The network consists of a sequence of two `tf.keras.layers.Dense` layers. These are densely connected, or fully connected, neural layers. The first `Dense` layer has 128 nodes (or neurons). The second (and last) layer returns a logits array with length of 5. Each node contains a score that indicates the current image belongs to one of the 5 classes.\n",
    "\n",
    "### Compile the model\n",
    "\n",
    "Before the model is ready for training, it needs a few more settings. These are added during the model's *compile* step:\n",
    "\n",
    "* *Loss function* —This measures how accurate the model is during training. You want to minimize this function to \"steer\" the model in the right direction.\n",
    "* *Optimizer* —This is how the model is updated based on the data it sees and its loss function.\n",
    "* *Metrics* —Used to monitor the training and testing steps. The following example uses *accuracy*, the fraction of the images that are correctly classified."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "executionInfo": {
     "elapsed": 1281,
     "status": "ok",
     "timestamp": 1618861397832,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "Lhan11blCaW7"
   },
   "outputs": [],
   "source": [
    "model.compile(optimizer='adam',\n",
    "              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1285,
     "status": "ok",
     "timestamp": 1618861398998,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "ym1DYZ3v0x5A",
    "outputId": "5b115ef5-a73b-4fef-8b96-b80497661fa1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_5\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "normalization_3 (Normalizati (None, 10)                21        \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 128)               1408      \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 5)                 645       \n",
      "=================================================================\n",
      "Total params: 2,074\n",
      "Trainable params: 2,053\n",
      "Non-trainable params: 21\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "qKF6uW-BCaW-"
   },
   "source": [
    "## Train the model\n",
    "\n",
    "Training the neural network model requires the following steps:\n",
    "\n",
    "1. Feed the training data to the model. In this example, the training data is in the `train_images` and `train_labels` arrays.\n",
    "2. The model learns to associate images and labels.\n",
    "3. You ask the model to make predictions about a test set—in this example, the `test_images` array.\n",
    "4. Verify that the predictions match the labels from the `test_labels` array.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Z4P4zIV7E28Z"
   },
   "source": [
    "### Feed the model\n",
    "\n",
    "To start training,  call the `model.fit` method—so called because it \"fits\" the model to the training data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 44143,
     "status": "ok",
     "timestamp": 1618861443814,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "xvwvpA64CaW_",
    "outputId": "cdcd6703-52cc-4b38-bcd5-7d725035fb47"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "98/98 [==============================] - 1s 4ms/step - loss: 1.8549 - accuracy: 0.2702 - val_loss: 1.3879 - val_accuracy: 0.3865\n",
      "Epoch 2/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.5215 - accuracy: 0.3620 - val_loss: 1.3430 - val_accuracy: 0.4365\n",
      "Epoch 3/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.4507 - accuracy: 0.3788 - val_loss: 1.2912 - val_accuracy: 0.4558\n",
      "Epoch 4/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.4419 - accuracy: 0.3768 - val_loss: 1.2512 - val_accuracy: 0.5173\n",
      "Epoch 5/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.3494 - accuracy: 0.4305 - val_loss: 1.2390 - val_accuracy: 0.4942\n",
      "Epoch 6/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.2862 - accuracy: 0.4691 - val_loss: 1.1962 - val_accuracy: 0.5404\n",
      "Epoch 7/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.2596 - accuracy: 0.4845 - val_loss: 1.1650 - val_accuracy: 0.5712\n",
      "Epoch 8/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.2541 - accuracy: 0.5016 - val_loss: 1.1309 - val_accuracy: 0.5692\n",
      "Epoch 9/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.1691 - accuracy: 0.5548 - val_loss: 1.1227 - val_accuracy: 0.5885\n",
      "Epoch 10/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.1581 - accuracy: 0.5551 - val_loss: 1.0900 - val_accuracy: 0.5865\n",
      "Epoch 11/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.1401 - accuracy: 0.5384 - val_loss: 1.0774 - val_accuracy: 0.5731\n",
      "Epoch 12/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.1348 - accuracy: 0.5648 - val_loss: 1.0497 - val_accuracy: 0.5846\n",
      "Epoch 13/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.0976 - accuracy: 0.5949 - val_loss: 1.0385 - val_accuracy: 0.6019\n",
      "Epoch 14/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.0713 - accuracy: 0.5965 - val_loss: 1.0268 - val_accuracy: 0.6135\n",
      "Epoch 15/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.0490 - accuracy: 0.6078 - val_loss: 1.0132 - val_accuracy: 0.6135\n",
      "Epoch 16/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.0446 - accuracy: 0.5956 - val_loss: 0.9863 - val_accuracy: 0.6385\n",
      "Epoch 17/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.0179 - accuracy: 0.6062 - val_loss: 0.9833 - val_accuracy: 0.6269\n",
      "Epoch 18/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 1.0083 - accuracy: 0.6237 - val_loss: 0.9585 - val_accuracy: 0.6442\n",
      "Epoch 19/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9950 - accuracy: 0.6365 - val_loss: 0.9590 - val_accuracy: 0.6481\n",
      "Epoch 20/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9715 - accuracy: 0.6376 - val_loss: 0.9356 - val_accuracy: 0.6519\n",
      "Epoch 21/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9699 - accuracy: 0.6300 - val_loss: 0.9325 - val_accuracy: 0.6615\n",
      "Epoch 22/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9629 - accuracy: 0.6495 - val_loss: 0.9320 - val_accuracy: 0.6731\n",
      "Epoch 23/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9356 - accuracy: 0.6668 - val_loss: 0.9111 - val_accuracy: 0.6808\n",
      "Epoch 24/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9326 - accuracy: 0.6652 - val_loss: 0.9037 - val_accuracy: 0.6827\n",
      "Epoch 25/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9298 - accuracy: 0.6510 - val_loss: 0.8909 - val_accuracy: 0.6750\n",
      "Epoch 26/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9096 - accuracy: 0.6685 - val_loss: 0.8938 - val_accuracy: 0.6962\n",
      "Epoch 27/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8954 - accuracy: 0.7010 - val_loss: 0.8903 - val_accuracy: 0.6846\n",
      "Epoch 28/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.9177 - accuracy: 0.6729 - val_loss: 0.8740 - val_accuracy: 0.6827\n",
      "Epoch 29/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8575 - accuracy: 0.6971 - val_loss: 0.8717 - val_accuracy: 0.7154\n",
      "Epoch 30/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8754 - accuracy: 0.6878 - val_loss: 0.8647 - val_accuracy: 0.6962\n",
      "Epoch 31/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8645 - accuracy: 0.6887 - val_loss: 0.8544 - val_accuracy: 0.7038\n",
      "Epoch 32/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8635 - accuracy: 0.6909 - val_loss: 0.8595 - val_accuracy: 0.6981\n",
      "Epoch 33/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8344 - accuracy: 0.7049 - val_loss: 0.8475 - val_accuracy: 0.7173\n",
      "Epoch 34/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7965 - accuracy: 0.7388 - val_loss: 0.8382 - val_accuracy: 0.6885\n",
      "Epoch 35/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8558 - accuracy: 0.6974 - val_loss: 0.8311 - val_accuracy: 0.7135\n",
      "Epoch 36/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8133 - accuracy: 0.7032 - val_loss: 0.8285 - val_accuracy: 0.7231\n",
      "Epoch 37/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8482 - accuracy: 0.7072 - val_loss: 0.8231 - val_accuracy: 0.7115\n",
      "Epoch 38/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8476 - accuracy: 0.6985 - val_loss: 0.8210 - val_accuracy: 0.7250\n",
      "Epoch 39/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8194 - accuracy: 0.6926 - val_loss: 0.8365 - val_accuracy: 0.7115\n",
      "Epoch 40/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7889 - accuracy: 0.7238 - val_loss: 0.8133 - val_accuracy: 0.7173\n",
      "Epoch 41/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8137 - accuracy: 0.6914 - val_loss: 0.8236 - val_accuracy: 0.6962\n",
      "Epoch 42/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7941 - accuracy: 0.7224 - val_loss: 0.8047 - val_accuracy: 0.7346\n",
      "Epoch 43/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8500 - accuracy: 0.6994 - val_loss: 0.7987 - val_accuracy: 0.7077\n",
      "Epoch 44/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7814 - accuracy: 0.7332 - val_loss: 0.8004 - val_accuracy: 0.7154\n",
      "Epoch 45/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.8015 - accuracy: 0.7039 - val_loss: 0.7938 - val_accuracy: 0.7173\n",
      "Epoch 46/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7738 - accuracy: 0.7291 - val_loss: 0.7998 - val_accuracy: 0.7058\n",
      "Epoch 47/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7917 - accuracy: 0.7129 - val_loss: 0.7897 - val_accuracy: 0.7212\n",
      "Epoch 48/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7550 - accuracy: 0.7412 - val_loss: 0.7828 - val_accuracy: 0.7288\n",
      "Epoch 49/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7551 - accuracy: 0.7445 - val_loss: 0.7936 - val_accuracy: 0.7288\n",
      "Epoch 50/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7342 - accuracy: 0.7449 - val_loss: 0.7988 - val_accuracy: 0.7231\n",
      "Epoch 51/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7770 - accuracy: 0.7305 - val_loss: 0.7902 - val_accuracy: 0.7115\n",
      "Epoch 52/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7622 - accuracy: 0.7286 - val_loss: 0.7815 - val_accuracy: 0.7115\n",
      "Epoch 53/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7479 - accuracy: 0.7433 - val_loss: 0.7718 - val_accuracy: 0.7308\n",
      "Epoch 54/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7601 - accuracy: 0.7326 - val_loss: 0.7758 - val_accuracy: 0.7135\n",
      "Epoch 55/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7482 - accuracy: 0.7304 - val_loss: 0.7772 - val_accuracy: 0.7115\n",
      "Epoch 56/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7244 - accuracy: 0.7408 - val_loss: 0.7645 - val_accuracy: 0.7154\n",
      "Epoch 57/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7281 - accuracy: 0.7209 - val_loss: 0.7828 - val_accuracy: 0.7077\n",
      "Epoch 58/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7702 - accuracy: 0.7325 - val_loss: 0.7668 - val_accuracy: 0.7173\n",
      "Epoch 59/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7191 - accuracy: 0.7443 - val_loss: 0.7591 - val_accuracy: 0.7173\n",
      "Epoch 60/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6780 - accuracy: 0.7536 - val_loss: 0.7579 - val_accuracy: 0.7288\n",
      "Epoch 61/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7097 - accuracy: 0.7394 - val_loss: 0.7622 - val_accuracy: 0.7327\n",
      "Epoch 62/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7462 - accuracy: 0.7265 - val_loss: 0.7587 - val_accuracy: 0.7327\n",
      "Epoch 63/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7447 - accuracy: 0.7361 - val_loss: 0.7499 - val_accuracy: 0.7308\n",
      "Epoch 64/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7401 - accuracy: 0.7401 - val_loss: 0.7597 - val_accuracy: 0.7346\n",
      "Epoch 65/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7384 - accuracy: 0.7327 - val_loss: 0.7453 - val_accuracy: 0.7308\n",
      "Epoch 66/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7053 - accuracy: 0.7589 - val_loss: 0.7621 - val_accuracy: 0.7269\n",
      "Epoch 67/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7289 - accuracy: 0.7550 - val_loss: 0.7519 - val_accuracy: 0.7231\n",
      "Epoch 68/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7545 - accuracy: 0.7282 - val_loss: 0.7545 - val_accuracy: 0.7327\n",
      "Epoch 69/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7253 - accuracy: 0.7409 - val_loss: 0.7425 - val_accuracy: 0.7365\n",
      "Epoch 70/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7190 - accuracy: 0.7445 - val_loss: 0.7377 - val_accuracy: 0.7404\n",
      "Epoch 71/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7114 - accuracy: 0.7551 - val_loss: 0.7470 - val_accuracy: 0.7404\n",
      "Epoch 72/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6811 - accuracy: 0.7530 - val_loss: 0.7458 - val_accuracy: 0.7269\n",
      "Epoch 73/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6945 - accuracy: 0.7365 - val_loss: 0.7501 - val_accuracy: 0.7346\n",
      "Epoch 74/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7087 - accuracy: 0.7471 - val_loss: 0.7618 - val_accuracy: 0.7250\n",
      "Epoch 75/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7285 - accuracy: 0.7410 - val_loss: 0.7451 - val_accuracy: 0.7327\n",
      "Epoch 76/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6644 - accuracy: 0.7516 - val_loss: 0.7394 - val_accuracy: 0.7250\n",
      "Epoch 77/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7315 - accuracy: 0.7356 - val_loss: 0.7448 - val_accuracy: 0.7346\n",
      "Epoch 78/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6932 - accuracy: 0.7423 - val_loss: 0.7341 - val_accuracy: 0.7288\n",
      "Epoch 79/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6685 - accuracy: 0.7558 - val_loss: 0.7339 - val_accuracy: 0.7385\n",
      "Epoch 80/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7121 - accuracy: 0.7505 - val_loss: 0.7357 - val_accuracy: 0.7365\n",
      "Epoch 81/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6903 - accuracy: 0.7528 - val_loss: 0.7373 - val_accuracy: 0.7385\n",
      "Epoch 82/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6952 - accuracy: 0.7643 - val_loss: 0.7221 - val_accuracy: 0.7250\n",
      "Epoch 83/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6815 - accuracy: 0.7461 - val_loss: 0.7264 - val_accuracy: 0.7327\n",
      "Epoch 84/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7003 - accuracy: 0.7509 - val_loss: 0.7430 - val_accuracy: 0.7077\n",
      "Epoch 85/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7003 - accuracy: 0.7389 - val_loss: 0.7274 - val_accuracy: 0.7500\n",
      "Epoch 86/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.7097 - accuracy: 0.7540 - val_loss: 0.7205 - val_accuracy: 0.7481\n",
      "Epoch 87/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6744 - accuracy: 0.7599 - val_loss: 0.7231 - val_accuracy: 0.7385\n",
      "Epoch 88/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6992 - accuracy: 0.7500 - val_loss: 0.7310 - val_accuracy: 0.7288\n",
      "Epoch 89/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6424 - accuracy: 0.7602 - val_loss: 0.7191 - val_accuracy: 0.7442\n",
      "Epoch 90/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6649 - accuracy: 0.7673 - val_loss: 0.7339 - val_accuracy: 0.7558\n",
      "Epoch 91/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6421 - accuracy: 0.7653 - val_loss: 0.7179 - val_accuracy: 0.7462\n",
      "Epoch 92/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6525 - accuracy: 0.7672 - val_loss: 0.7321 - val_accuracy: 0.7442\n",
      "Epoch 93/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6564 - accuracy: 0.7612 - val_loss: 0.7238 - val_accuracy: 0.7481\n",
      "Epoch 94/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6786 - accuracy: 0.7438 - val_loss: 0.7255 - val_accuracy: 0.7308\n",
      "Epoch 95/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6830 - accuracy: 0.7570 - val_loss: 0.7078 - val_accuracy: 0.7404\n",
      "Epoch 96/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6501 - accuracy: 0.7445 - val_loss: 0.7138 - val_accuracy: 0.7404\n",
      "Epoch 97/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6757 - accuracy: 0.7597 - val_loss: 0.7114 - val_accuracy: 0.7442\n",
      "Epoch 98/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6689 - accuracy: 0.7580 - val_loss: 0.7183 - val_accuracy: 0.7462\n",
      "Epoch 99/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6659 - accuracy: 0.7605 - val_loss: 0.7210 - val_accuracy: 0.7404\n",
      "Epoch 100/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6742 - accuracy: 0.7608 - val_loss: 0.7261 - val_accuracy: 0.7577\n",
      "Epoch 101/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6950 - accuracy: 0.7496 - val_loss: 0.7269 - val_accuracy: 0.7519\n",
      "Epoch 102/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6784 - accuracy: 0.7484 - val_loss: 0.7170 - val_accuracy: 0.7519\n",
      "Epoch 103/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6616 - accuracy: 0.7572 - val_loss: 0.6997 - val_accuracy: 0.7442\n",
      "Epoch 104/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6590 - accuracy: 0.7606 - val_loss: 0.7043 - val_accuracy: 0.7385\n",
      "Epoch 105/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6587 - accuracy: 0.7594 - val_loss: 0.7072 - val_accuracy: 0.7423\n",
      "Epoch 106/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6885 - accuracy: 0.7425 - val_loss: 0.7005 - val_accuracy: 0.7442\n",
      "Epoch 107/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6454 - accuracy: 0.7743 - val_loss: 0.7091 - val_accuracy: 0.7635\n",
      "Epoch 108/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6602 - accuracy: 0.7705 - val_loss: 0.7016 - val_accuracy: 0.7442\n",
      "Epoch 109/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6513 - accuracy: 0.7529 - val_loss: 0.7029 - val_accuracy: 0.7481\n",
      "Epoch 110/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6325 - accuracy: 0.7784 - val_loss: 0.7026 - val_accuracy: 0.7577\n",
      "Epoch 111/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6657 - accuracy: 0.7690 - val_loss: 0.7026 - val_accuracy: 0.7481\n",
      "Epoch 112/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6701 - accuracy: 0.7554 - val_loss: 0.7053 - val_accuracy: 0.7538\n",
      "Epoch 113/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6511 - accuracy: 0.7808 - val_loss: 0.7025 - val_accuracy: 0.7558\n",
      "Epoch 114/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6750 - accuracy: 0.7618 - val_loss: 0.6922 - val_accuracy: 0.7346\n",
      "Epoch 115/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6360 - accuracy: 0.7779 - val_loss: 0.7014 - val_accuracy: 0.7538\n",
      "Epoch 116/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6520 - accuracy: 0.7605 - val_loss: 0.7014 - val_accuracy: 0.7673\n",
      "Epoch 117/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6078 - accuracy: 0.7874 - val_loss: 0.7044 - val_accuracy: 0.7365\n",
      "Epoch 118/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6259 - accuracy: 0.7764 - val_loss: 0.6874 - val_accuracy: 0.7615\n",
      "Epoch 119/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6577 - accuracy: 0.7605 - val_loss: 0.6984 - val_accuracy: 0.7558\n",
      "Epoch 120/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6419 - accuracy: 0.7645 - val_loss: 0.6886 - val_accuracy: 0.7615\n",
      "Epoch 121/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6369 - accuracy: 0.7669 - val_loss: 0.7017 - val_accuracy: 0.7558\n",
      "Epoch 122/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6486 - accuracy: 0.7620 - val_loss: 0.7003 - val_accuracy: 0.7577\n",
      "Epoch 123/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6188 - accuracy: 0.7654 - val_loss: 0.6918 - val_accuracy: 0.7538\n",
      "Epoch 124/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6407 - accuracy: 0.7716 - val_loss: 0.6984 - val_accuracy: 0.7635\n",
      "Epoch 125/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6219 - accuracy: 0.7885 - val_loss: 0.6920 - val_accuracy: 0.7596\n",
      "Epoch 126/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6555 - accuracy: 0.7531 - val_loss: 0.6906 - val_accuracy: 0.7654\n",
      "Epoch 127/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6242 - accuracy: 0.7669 - val_loss: 0.6969 - val_accuracy: 0.7577\n",
      "Epoch 128/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6361 - accuracy: 0.7687 - val_loss: 0.6910 - val_accuracy: 0.7654\n",
      "Epoch 129/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6572 - accuracy: 0.7571 - val_loss: 0.6835 - val_accuracy: 0.7558\n",
      "Epoch 130/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6313 - accuracy: 0.7610 - val_loss: 0.6935 - val_accuracy: 0.7481\n",
      "Epoch 131/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6238 - accuracy: 0.7836 - val_loss: 0.7052 - val_accuracy: 0.7500\n",
      "Epoch 132/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6299 - accuracy: 0.7602 - val_loss: 0.6986 - val_accuracy: 0.7500\n",
      "Epoch 133/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6566 - accuracy: 0.7672 - val_loss: 0.6949 - val_accuracy: 0.7596\n",
      "Epoch 134/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6286 - accuracy: 0.7665 - val_loss: 0.6938 - val_accuracy: 0.7577\n",
      "Epoch 135/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6259 - accuracy: 0.7738 - val_loss: 0.6906 - val_accuracy: 0.7519\n",
      "Epoch 136/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6141 - accuracy: 0.7708 - val_loss: 0.6853 - val_accuracy: 0.7654\n",
      "Epoch 137/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5885 - accuracy: 0.7849 - val_loss: 0.6952 - val_accuracy: 0.7615\n",
      "Epoch 138/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6171 - accuracy: 0.7741 - val_loss: 0.6927 - val_accuracy: 0.7596\n",
      "Epoch 139/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5820 - accuracy: 0.7901 - val_loss: 0.6879 - val_accuracy: 0.7673\n",
      "Epoch 140/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5959 - accuracy: 0.7781 - val_loss: 0.6947 - val_accuracy: 0.7442\n",
      "Epoch 141/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6180 - accuracy: 0.7858 - val_loss: 0.6857 - val_accuracy: 0.7654\n",
      "Epoch 142/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6247 - accuracy: 0.7778 - val_loss: 0.6838 - val_accuracy: 0.7538\n",
      "Epoch 143/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6107 - accuracy: 0.7822 - val_loss: 0.6845 - val_accuracy: 0.7635\n",
      "Epoch 144/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6013 - accuracy: 0.7928 - val_loss: 0.6772 - val_accuracy: 0.7654\n",
      "Epoch 145/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6646 - accuracy: 0.7628 - val_loss: 0.6861 - val_accuracy: 0.7519\n",
      "Epoch 146/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6176 - accuracy: 0.7761 - val_loss: 0.6809 - val_accuracy: 0.7654\n",
      "Epoch 147/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6158 - accuracy: 0.7695 - val_loss: 0.6832 - val_accuracy: 0.7673\n",
      "Epoch 148/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6139 - accuracy: 0.7677 - val_loss: 0.6761 - val_accuracy: 0.7673\n",
      "Epoch 149/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6270 - accuracy: 0.7792 - val_loss: 0.6920 - val_accuracy: 0.7442\n",
      "Epoch 150/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6103 - accuracy: 0.7842 - val_loss: 0.6851 - val_accuracy: 0.7519\n",
      "Epoch 151/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6312 - accuracy: 0.7598 - val_loss: 0.6732 - val_accuracy: 0.7635\n",
      "Epoch 152/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6420 - accuracy: 0.7716 - val_loss: 0.6790 - val_accuracy: 0.7615\n",
      "Epoch 153/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6009 - accuracy: 0.7968 - val_loss: 0.6898 - val_accuracy: 0.7558\n",
      "Epoch 154/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6372 - accuracy: 0.7653 - val_loss: 0.6769 - val_accuracy: 0.7635\n",
      "Epoch 155/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6253 - accuracy: 0.7699 - val_loss: 0.7003 - val_accuracy: 0.7654\n",
      "Epoch 156/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6425 - accuracy: 0.7749 - val_loss: 0.6783 - val_accuracy: 0.7673\n",
      "Epoch 157/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5940 - accuracy: 0.7870 - val_loss: 0.6701 - val_accuracy: 0.7712\n",
      "Epoch 158/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5924 - accuracy: 0.7883 - val_loss: 0.6937 - val_accuracy: 0.7462\n",
      "Epoch 159/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6223 - accuracy: 0.7900 - val_loss: 0.6705 - val_accuracy: 0.7577\n",
      "Epoch 160/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6250 - accuracy: 0.7762 - val_loss: 0.6768 - val_accuracy: 0.7673\n",
      "Epoch 161/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5967 - accuracy: 0.7922 - val_loss: 0.6675 - val_accuracy: 0.7731\n",
      "Epoch 162/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5739 - accuracy: 0.7949 - val_loss: 0.6778 - val_accuracy: 0.7731\n",
      "Epoch 163/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6516 - accuracy: 0.7442 - val_loss: 0.6702 - val_accuracy: 0.7712\n",
      "Epoch 164/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6020 - accuracy: 0.7777 - val_loss: 0.6709 - val_accuracy: 0.7654\n",
      "Epoch 165/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5989 - accuracy: 0.7820 - val_loss: 0.6815 - val_accuracy: 0.7577\n",
      "Epoch 166/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6047 - accuracy: 0.7844 - val_loss: 0.6641 - val_accuracy: 0.7615\n",
      "Epoch 167/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6180 - accuracy: 0.7730 - val_loss: 0.6686 - val_accuracy: 0.7712\n",
      "Epoch 168/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5752 - accuracy: 0.7907 - val_loss: 0.6683 - val_accuracy: 0.7750\n",
      "Epoch 169/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5827 - accuracy: 0.7849 - val_loss: 0.6798 - val_accuracy: 0.7673\n",
      "Epoch 170/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6310 - accuracy: 0.7768 - val_loss: 0.6624 - val_accuracy: 0.7673\n",
      "Epoch 171/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5964 - accuracy: 0.7885 - val_loss: 0.6647 - val_accuracy: 0.7654\n",
      "Epoch 172/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6048 - accuracy: 0.7876 - val_loss: 0.6865 - val_accuracy: 0.7673\n",
      "Epoch 173/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5942 - accuracy: 0.7906 - val_loss: 0.6681 - val_accuracy: 0.7615\n",
      "Epoch 174/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5657 - accuracy: 0.7816 - val_loss: 0.6697 - val_accuracy: 0.7673\n",
      "Epoch 175/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6114 - accuracy: 0.7781 - val_loss: 0.6653 - val_accuracy: 0.7635\n",
      "Epoch 176/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5779 - accuracy: 0.8125 - val_loss: 0.6808 - val_accuracy: 0.7673\n",
      "Epoch 177/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5815 - accuracy: 0.7939 - val_loss: 0.6725 - val_accuracy: 0.7692\n",
      "Epoch 178/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5786 - accuracy: 0.7942 - val_loss: 0.6723 - val_accuracy: 0.7654\n",
      "Epoch 179/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6160 - accuracy: 0.7792 - val_loss: 0.6617 - val_accuracy: 0.7519\n",
      "Epoch 180/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5538 - accuracy: 0.8034 - val_loss: 0.6654 - val_accuracy: 0.7673\n",
      "Epoch 181/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6256 - accuracy: 0.7702 - val_loss: 0.6623 - val_accuracy: 0.7673\n",
      "Epoch 182/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5722 - accuracy: 0.7956 - val_loss: 0.6775 - val_accuracy: 0.7673\n",
      "Epoch 183/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6127 - accuracy: 0.7825 - val_loss: 0.6735 - val_accuracy: 0.7673\n",
      "Epoch 184/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5505 - accuracy: 0.8025 - val_loss: 0.6634 - val_accuracy: 0.7692\n",
      "Epoch 185/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5836 - accuracy: 0.7843 - val_loss: 0.6727 - val_accuracy: 0.7654\n",
      "Epoch 186/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6168 - accuracy: 0.7723 - val_loss: 0.6640 - val_accuracy: 0.7558\n",
      "Epoch 187/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5749 - accuracy: 0.7995 - val_loss: 0.6630 - val_accuracy: 0.7635\n",
      "Epoch 188/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5411 - accuracy: 0.8099 - val_loss: 0.6582 - val_accuracy: 0.7769\n",
      "Epoch 189/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5902 - accuracy: 0.7868 - val_loss: 0.6970 - val_accuracy: 0.7596\n",
      "Epoch 190/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5631 - accuracy: 0.7806 - val_loss: 0.6687 - val_accuracy: 0.7769\n",
      "Epoch 191/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.6292 - accuracy: 0.7633 - val_loss: 0.6710 - val_accuracy: 0.7692\n",
      "Epoch 192/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5737 - accuracy: 0.8003 - val_loss: 0.6691 - val_accuracy: 0.7673\n",
      "Epoch 193/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5517 - accuracy: 0.8069 - val_loss: 0.6680 - val_accuracy: 0.7558\n",
      "Epoch 194/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5639 - accuracy: 0.7845 - val_loss: 0.6668 - val_accuracy: 0.7673\n",
      "Epoch 195/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5997 - accuracy: 0.8004 - val_loss: 0.6574 - val_accuracy: 0.7673\n",
      "Epoch 196/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5739 - accuracy: 0.8073 - val_loss: 0.6604 - val_accuracy: 0.7712\n",
      "Epoch 197/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5848 - accuracy: 0.7926 - val_loss: 0.6608 - val_accuracy: 0.7596\n",
      "Epoch 198/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5596 - accuracy: 0.7939 - val_loss: 0.6577 - val_accuracy: 0.7769\n",
      "Epoch 199/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5643 - accuracy: 0.7978 - val_loss: 0.6603 - val_accuracy: 0.7788\n",
      "Epoch 200/200\n",
      "98/98 [==============================] - 0s 2ms/step - loss: 0.5555 - accuracy: 0.8046 - val_loss: 0.6593 - val_accuracy: 0.7769\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(\n",
    "    x=train_features,\n",
    "    y=train_labels,\n",
    "    batch_size=16,\n",
    "    validation_data=(test_features,test_labels),\n",
    "    epochs=200,\n",
    "    verbose=1,\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "W3ZVOhugCaXA"
   },
   "source": [
    "As the model trains, the loss and accuracy metrics are displayed. This model reaches an accuracy of about 80% on the training data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "wCpr6DGyE28h"
   },
   "source": [
    "### Evaluate accuracy\n",
    "\n",
    "Next, compare how the model performs on the test dataset:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1641,
     "status": "ok",
     "timestamp": 1618859329477,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "t6FlVCCS7NeL",
    "outputId": "4cb1f512-5245-4406-a504-521ff4ca6f04"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])"
      ]
     },
     "execution_count": 23,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dict.keys(history.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 268
    },
    "executionInfo": {
     "elapsed": 1130,
     "status": "ok",
     "timestamp": 1618861450193,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "VflXLEeECaXC",
    "outputId": "00faf558-5c00-44aa-e230-3f690dad70c9"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "metrics = history.history\n",
    "plt.plot(history.epoch, metrics['loss'])\n",
    "plt.plot(history.epoch, metrics['accuracy'])\n",
    "plt.plot(history.epoch, metrics['val_loss'])\n",
    "plt.plot(history.epoch, metrics['val_accuracy'])\n",
    "plt.legend(['loss', 'accuracy', 'val_loss', 'val_accuracy'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "yWfgsmVXCaXG"
   },
   "source": [
    "It turns out that the accuracy on the test dataset is a little less than the accuracy on the training dataset. This gap between training accuracy and test accuracy represents *overfitting*. Overfitting happens when a machine learning model performs worse on new, previously unseen inputs than it does on the training data. An overfitted model \"memorizes\" the noise and details in the training dataset to a point where it negatively impacts the performance of the model on the new data. \n",
    "\n",
    "One way to deal with this is by introducing some regularization with a [Dropout](https://jmlr.org/papers/volume15/srivastava14a.old/srivastava14a.pdf) layer. To do this, you need to go back to re-initialize the `Sequential` class with one layer being `tk.keras.layers.Dropout(0.2)`, where `0.2` is the amount of units in percentage to remove between layers. The more you take, the stronger the regularization."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "v-PyD1SYE28q"
   },
   "source": [
    "### Make predictions\n",
    "\n",
    "With the model trained, you can use it to make predictions about some images.\n",
    "The model's linear outputs, [logits](https://developers.google.com/machine-learning/glossary#logits). Attach a softmax layer to convert the logits to probabilities, which are easier to interpret. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "executionInfo": {
     "elapsed": 1108,
     "status": "ok",
     "timestamp": 1618861468393,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "DnfNA0CrQLSD"
   },
   "outputs": [],
   "source": [
    "probability_model = tf.keras.Sequential([model, tf.keras.layers.Softmax()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "executionInfo": {
     "elapsed": 1263,
     "status": "ok",
     "timestamp": 1618861469078,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "Gl91RPhdCaXI"
   },
   "outputs": [],
   "source": [
    "predictions = probability_model.predict(test_features)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "x9Kk1voUCaXJ"
   },
   "source": [
    "Here, the model has predicted the label for each artist in the testing set. Let's take a look at the first prediction:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1415,
     "status": "ok",
     "timestamp": 1618860701718,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "3DmJEUinCaXK",
    "outputId": "4eb551e3-f0db-4b3d-8d74-729ee5e87b46"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1.1181573e-02, 9.8401791e-01, 9.3440511e-05, 9.6941723e-05,\n",
       "       4.6101473e-03], dtype=float32)"
      ]
     },
     "execution_count": 65,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-hw1hgeSCaXN"
   },
   "source": [
    "A prediction is an array of 5 numbers. They represent the model's \"confidence\" that the image corresponds to each of the 5 different articles of clothing. You can see which label has the highest confidence value:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1007,
     "status": "ok",
     "timestamp": 1618860702147,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "qsqenuPnCaXO",
    "outputId": "eade39c8-1167-4ab6-96a5-694cf5f2f157"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1"
      ]
     },
     "execution_count": 66,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.argmax(predictions[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "E51yS7iCCaXO"
   },
   "source": [
    "So, the model is most confident that this is artist `class_names[1]`. Examining the test label shows that this classification is correct:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 719,
     "status": "ok",
     "timestamp": 1618860703479,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "Sd7Pgsu6CaXP",
    "outputId": "ccd86124-124c-4b18-fde0-d20e92f21722"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0"
      ]
     },
     "execution_count": 67,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_labels[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "P9p3ephyASHw"
   },
   "source": [
    "## Confusion Matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1343,
     "status": "ok",
     "timestamp": 1618861471487,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "h6t5Fk0o3gG0",
    "outputId": "6215e56a-9ec8-4f3c-beab-e6335a3bcfa5"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((520,), (520,))"
      ]
     },
     "execution_count": 90,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_labels = np.array([np.argmax(p) for p in predictions])\n",
    "pred_labels.shape, test_labels.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 362
    },
    "executionInfo": {
     "elapsed": 1761,
     "status": "ok",
     "timestamp": 1618861472505,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "hQlnbqaw2Qu_",
    "outputId": "c7783422-ef97-4a8a-c4cf-23daabc36cdf"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "\n",
    "mat = tf.math.confusion_matrix(test_labels, pred_labels)\n",
    "\n",
    "sns.heatmap(mat, square=True, annot=True, cbar=True, fmt='d',\n",
    "            xticklabels=label_names,\n",
    "            yticklabels=label_names)\n",
    "\n",
    "plt.xlabel('True')\n",
    "plt.ylabel('Predicted');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "R32zteKHCaXT"
   },
   "source": [
    "## Use the trained model\n",
    "\n",
    "Finally, use the trained model to make a prediction about a single image."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "executionInfo": {
     "elapsed": 1291,
     "status": "ok",
     "timestamp": 1618860738891,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "Wadf0J6RBgct"
   },
   "outputs": [],
   "source": [
    "stats = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1371,
     "status": "ok",
     "timestamp": 1618861857718,
     "user": {
      "displayName": "Fede Camara Halac",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14Gh_b6Y4JRn1pQwmH2PwmvUr8HiKg-a_TZFj1JfXLQ=s64",
      "userId": "14838610272939224163"
     },
     "user_tz": 240
    },
    "id": "yRJ7JU7JCaXT",
    "outputId": "8aec39b3-241d-47e6-a43b-7d55d16523a9"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[32m================================================================================\u001b[0m\n",
      "\u001b[32m| Knew it.\u001b[0m\n",
      "\u001b[32m--------------------------------------------------------------------------------\u001b[0m\n",
      "\u001b[32m| TRUE: The Rolling Stones\u001b[0m\n",
      "\u001b[32m| PRED: The Rolling Stones\u001b[0m\n",
      "\u001b[32m================================================================================\u001b[0m\n",
      "Current accuracy: 76.92307692307693\n",
      "Actual accuracy is: 79.67948913574219\n"
     ]
    }
   ],
   "source": [
    "# Grab a random element from the test dataset.\n",
    "example = np.random.randint(len(test_features))\n",
    "\n",
    "tester_feature = test_features[example]\n",
    "\n",
    "# `tf.keras` models are optimized to make predictions on a *batch*, \n",
    "# or collection, of examples at once. Accordingly, even though \n",
    "# you're using a single image, you need to add it to a list:\n",
    "\n",
    "# Add the image to a batch where it's the only member.\n",
    "tester_feature = (np.expand_dims(tester_feature,0))\n",
    "\n",
    "\n",
    "# Now predict the correct label for this image:\n",
    "predictions_single = probability_model.predict(tester_feature)\n",
    "\n",
    "y_pred = np.argmax(predictions_single[0])\n",
    "y_true = int(test_labels[example])\n",
    "\n",
    "# get both Predicted and True labels:\n",
    "\n",
    "from termcolor import colored\n",
    "\n",
    "\n",
    "\n",
    "if y_true != y_pred:\n",
    "    stats.append(0)\n",
    "    msg = f\"Yikes!, {['Oh, well', 'perhaps next time?', 'ugh...'][np.random.randint(3)]}\"\n",
    "    color='red'\n",
    "else:\n",
    "    stats.append(1)\n",
    "    msg = f\"Knew it.\"\n",
    "    color='green'\n",
    "\n",
    "print(colored(\"=\"*80, color))\n",
    "print(colored(\"| \"+msg, color))\n",
    "print(colored(\"-\"*80, color))\n",
    "print(colored(\"| TRUE: \"+label_names[y_true], color))\n",
    "print(colored(\"| PRED: \"+label_names[y_pred], color))\n",
    "print(colored(\"=\"*80, color))\n",
    "\n",
    "\n",
    "print(f\"Current accuracy: {stats.count(1) / len(stats) * 100}\")\n",
    "print(f\"Actual accuracy is: {metrics['accuracy'][-1]*100}\")\n"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "Week 15 - Neural Networks.ipynb",
   "provenance": [
    {
     "file_id": "1cQpg4tG6TegX0cMWuXaKlqTF-RtYCw78",
     "timestamp": 1618852691605
    },
    {
     "file_id": "https://github.com/tensorflow/docs/blob/master/site/en/tutorials/keras/classification.ipynb",
     "timestamp": 1618840508023
    }
   ]
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}