mnist-digit-recognigtion-0-9974-with-tensorflow.ipynb

{"cells":[{"source":"<a href=\"https://www.kaggle.com/code/peremartramanonellas/mnist-digit-recognigtion-0-9974-with-tensorflow?scriptVersionId=124698404\" target=\"_blank\"><img align=\"left\" alt=\"Kaggle\" title=\"Open in Kaggle\" src=\"https://kaggle.com/static/images/open-in-kaggle.svg\"></a>","metadata":{},"cell_type":"markdown"},{"cell_type":"code","execution_count":1,"id":"2711cb8c","metadata":{"_kg_hide-input":true,"execution":{"iopub.execute_input":"2023-04-05T14:21:26.321169Z","iopub.status.busy":"2023-04-05T14:21:26.320617Z","iopub.status.idle":"2023-04-05T14:21:26.335409Z","shell.execute_reply":"2023-04-05T14:21:26.334402Z"},"papermill":{"duration":0.02728,"end_time":"2023-04-05T14:21:26.338345","exception":false,"start_time":"2023-04-05T14:21:26.311065","status":"completed"},"tags":[]},"outputs":[{"data":{"text/html":["\n","<style>\n",".output_png {\n","    display: table-cell;\n","    text-align: center;\n","    vertical-align: middle;\n","    horizontal-align: middle;\n","}\n","h1 {\n","    text-align: center;\n","    background-color: AliceBlue;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","h2 {\n","    text-align: center;\n","    background-color: HoneyDew;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","h3 {\n","    text-align: center;\n","    background-color: MintCream;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","\n","body, p {\n","    font-family: monospace;\n","    font-size: 15px;\n","    color: charcoal;\n","}\n","div {\n","    font-size: 14px;\n","    margin: 0;\n","\n","}\n","\n","h4 {\n","    padding: 0px;\n","    margin: 0;\n","    font-family: monospace;\n","    color: purple;\n","}\n","</style>\n"],"text/plain":["<IPython.core.display.HTML object>"]},"execution_count":1,"metadata":{},"output_type":"execute_result"}],"source":["from IPython.core.display import HTML\n","HTML(\"\"\"\n","<style>\n",".output_png {\n","    display: table-cell;\n","    text-align: center;\n","    vertical-align: middle;\n","    horizontal-align: middle;\n","}\n","h1 {\n","    text-align: center;\n","    background-color: AliceBlue;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","h2 {\n","    text-align: center;\n","    background-color: HoneyDew;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","h3 {\n","    text-align: center;\n","    background-color: MintCream;\n","    padding: 10px;\n","    margin: 0;\n","    font-family: monospace;\n","    color:DimGray;\n","    border-radius: 10px\n","}\n","\n","\n","body, p {\n","    font-family: monospace;\n","    font-size: 15px;\n","    color: charcoal;\n","}\n","div {\n","    font-size: 14px;\n","    margin: 0;\n","\n","}\n","\n","h4 {\n","    padding: 0px;\n","    margin: 0;\n","    font-family: monospace;\n","    color: purple;\n","}\n","</style>\n","\"\"\")"]},{"cell_type":"markdown","id":"89584727","metadata":{"papermill":{"duration":0.006168,"end_time":"2023-04-05T14:21:26.351072","exception":false,"start_time":"2023-04-05T14:21:26.344904","status":"completed"},"tags":[]},"source":["# MNIST using Tensorflow. 0.996 in VAL_Accuracy \n","\n","Mi first submission to Kaggle. I achieved a VAL_ACCURACY between 0.9955 and 0.9972 with 30 epochs.  \n","\n","The model is really common, with a mix of Conv2D and BatchNormalization layers. \n","I used a **SpatialDropout2D** layer instead of a Dropout. The difference between this two layers is explained below. \n","\n","I used a **bit of ImageAugmentaion**, I don't know how the test data is, but the images are realy simple, and it's totally unnecesary use vertical or horizontal rotations, just some zoom and shift. \n","\n","**Feel free to Copy & Edit, or fork this Notebook and use to do a better one, jus please upvote the notebbok if you like and use it. **"]},{"cell_type":"code","execution_count":2,"id":"1e527835","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:26.365127Z","iopub.status.busy":"2023-04-05T14:21:26.364567Z","iopub.status.idle":"2023-04-05T14:21:32.244988Z","shell.execute_reply":"2023-04-05T14:21:32.244052Z"},"papermill":{"duration":5.890168,"end_time":"2023-04-05T14:21:32.247505","exception":false,"start_time":"2023-04-05T14:21:26.357337","status":"completed"},"tags":[]},"outputs":[],"source":["import tensorflow as tf\n","tf.random.set_seed(42)\n","import numpy as np\n","\n","from tensorflow import keras\n","from tensorflow.keras.models import load_model\n","from time import time\n","import matplotlib.pyplot as plt\n","import matplotlib.image as mpimg\n","from matplotlib import rcParams\n","import pandas as pd\n","from sklearn.model_selection import train_test_split"]},{"cell_type":"markdown","id":"0c66f347","metadata":{"papermill":{"duration":0.006133,"end_time":"2023-04-05T14:21:32.260298","exception":false,"start_time":"2023-04-05T14:21:32.254165","status":"completed"},"tags":[]},"source":["# SUPPORT FUNCTIONS\n","Some functions that I use in a lot of different Notebooks. \n"]},{"cell_type":"markdown","id":"6449f8c2","metadata":{"papermill":{"duration":0.006007,"end_time":"2023-04-05T14:21:32.272532","exception":false,"start_time":"2023-04-05T14:21:32.266525","status":"completed"},"tags":[]},"source":["**plot_loss_acc(history, firstepoch=0)**\n","\n","I Use this function to plot the loss and accuracy of the training and validation data of the history received. You can indicate the first epoch to plot. \n","\n","Useful to see the slope of the curve in the last epochs. \n","\n","**Parameteres:**\n","* history--> The history returned by the fit of the model.\n","* firstepoch--> First epoch to show in the plot. Is useful, because the first epoch should be a really bad result, and it's difficult to apreciate the curve of the line in the last epochs. "]},{"cell_type":"code","execution_count":3,"id":"4244dfa8","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:32.288043Z","iopub.status.busy":"2023-04-05T14:21:32.28618Z","iopub.status.idle":"2023-04-05T14:21:32.294671Z","shell.execute_reply":"2023-04-05T14:21:32.2938Z"},"papermill":{"duration":0.017776,"end_time":"2023-04-05T14:21:32.296575","exception":false,"start_time":"2023-04-05T14:21:32.278799","status":"completed"},"tags":[]},"outputs":[],"source":["\n","def plot_loss_acc(history, firstepoch=0):\n","  '''Plots the training and validation loss and accuracy from a history object'''\n","  acc = history.history['accuracy']\n","  acc = acc[firstepoch:]\n","  val_acc = history.history['val_accuracy']\n","    \n","  val_acc = val_acc[firstepoch:]\n","  loss = history.history['loss']\n","  loss=loss[firstepoch:]\n","  val_loss = history.history['val_loss']\n","  val_loss = val_loss[firstepoch:]\n","\n","  epochs = range(len(acc))\n","\n","  plt.plot(epochs, acc, 'bo-', label='Training accuracy')\n","  plt.plot(epochs, val_acc, 'go-', label='Validation accuracy')\n","  plt.title('Training and validation accuracy')\n","  plt.legend()\n","\n","  plt.figure()\n","\n","  plt.plot(epochs, loss, 'bo-', label='Training Loss')\n","  plt.plot(epochs, val_loss, 'go-', label='Validation Loss')\n","  plt.title('Training and validation loss')\n","  plt.legend()\n","\n","  plt.show()"]},{"cell_type":"markdown","id":"0777b7ce","metadata":{"papermill":{"duration":0.00598,"end_time":"2023-04-05T14:21:32.308696","exception":false,"start_time":"2023-04-05T14:21:32.302716","status":"completed"},"tags":[]},"source":["**show_img_dataset(X, y=None, nrows = 4, ncols=4, firstimg=100, numimg=4)**\n","\n","Show images from a dataset. It have a lot of parameters, but just one is obligatory. \n","\n","Parameters: \n","* X--> The dataset containing the images. \n","* y--> The labels. If informed it will appear below the image. \n","* nrows\n","* ncols\n","* firstimg--> The first img to show, if to big the function fails, I don't check it. \n","* numimg--> The number of images to display. \n"]},{"cell_type":"code","execution_count":4,"id":"7a2479db","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:32.324597Z","iopub.status.busy":"2023-04-05T14:21:32.322916Z","iopub.status.idle":"2023-04-05T14:21:32.329463Z","shell.execute_reply":"2023-04-05T14:21:32.328601Z"},"papermill":{"duration":0.015732,"end_time":"2023-04-05T14:21:32.331432","exception":false,"start_time":"2023-04-05T14:21:32.3157","status":"completed"},"tags":[]},"outputs":[],"source":["def show_img_dataset(X, y=None, nrows = 4, ncols=4, firstimg=100, numimg=4):\n","    for i in range(numimg):\n","        sp = plt.subplot(nrows, ncols, i + 1)\n","\n","        sp.axis('Off')\n","        plt.imshow(X[firstimg+i], cmap=\"Greys\")\n","        if (y is not None): \n","            plt.title(y[firstimg+i])\n","    plt.show()"]},{"cell_type":"code","execution_count":5,"id":"175fcb59","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:32.346476Z","iopub.status.busy":"2023-04-05T14:21:32.345048Z","iopub.status.idle":"2023-04-05T14:21:32.350874Z","shell.execute_reply":"2023-04-05T14:21:32.349995Z"},"papermill":{"duration":0.015001,"end_time":"2023-04-05T14:21:32.352785","exception":false,"start_time":"2023-04-05T14:21:32.337784","status":"completed"},"tags":[]},"outputs":[],"source":["#return predictions of the model received. \n","def get_predictions(model, X, y):\n","    predictions=model.predict(X)  \n","    results = pd.DataFrame(data={'Predictions': np.argmax(predictions, axis=1), \n","                            'Actuals': np.argmax(y, axis=1)})\n","    predictions.shape, y.shape\n","    return predictions, results"]},{"cell_type":"markdown","id":"0fd54aa6","metadata":{"papermill":{"duration":0.006035,"end_time":"2023-04-05T14:21:32.364981","exception":false,"start_time":"2023-04-05T14:21:32.358946","status":"completed"},"tags":[]},"source":["# Load and treatment of Data"]},{"cell_type":"code","execution_count":6,"id":"84a9c3fc","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:32.379445Z","iopub.status.busy":"2023-04-05T14:21:32.378551Z","iopub.status.idle":"2023-04-05T14:21:37.441Z","shell.execute_reply":"2023-04-05T14:21:37.44002Z"},"papermill":{"duration":5.072066,"end_time":"2023-04-05T14:21:37.443409","exception":false,"start_time":"2023-04-05T14:21:32.371343","status":"completed"},"tags":[]},"outputs":[],"source":["#Read the data from CSV files. \n","train_df=pd.read_csv('/kaggle/input/digit-recognizer/train.csv')\n","test_df=pd.read_csv('/kaggle/input/digit-recognizer/test.csv')"]},{"cell_type":"code","execution_count":7,"id":"c35b39ea","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:37.45878Z","iopub.status.busy":"2023-04-05T14:21:37.45847Z","iopub.status.idle":"2023-04-05T14:21:37.479816Z","shell.execute_reply":"2023-04-05T14:21:37.478922Z"},"papermill":{"duration":0.03171,"end_time":"2023-04-05T14:21:37.481839","exception":false,"start_time":"2023-04-05T14:21:37.450129","status":"completed"},"tags":[]},"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>label</th>\n","      <th>pixel0</th>\n","      <th>pixel1</th>\n","      <th>pixel2</th>\n","      <th>pixel3</th>\n","      <th>pixel4</th>\n","      <th>pixel5</th>\n","      <th>pixel6</th>\n","      <th>pixel7</th>\n","      <th>pixel8</th>\n","      <th>...</th>\n","      <th>pixel774</th>\n","      <th>pixel775</th>\n","      <th>pixel776</th>\n","      <th>pixel777</th>\n","      <th>pixel778</th>\n","      <th>pixel779</th>\n","      <th>pixel780</th>\n","      <th>pixel781</th>\n","      <th>pixel782</th>\n","      <th>pixel783</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>1</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>...</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>...</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>1</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>...</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>4</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>...</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>...</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","      <td>0</td>\n","    </tr>\n","  </tbody>\n","</table>\n","<p>5 rows × 785 columns</p>\n","</div>"],"text/plain":["   label  pixel0  pixel1  pixel2  pixel3  pixel4  pixel5  pixel6  pixel7  \\\n","0      1       0       0       0       0       0       0       0       0   \n","1      0       0       0       0       0       0       0       0       0   \n","2      1       0       0       0       0       0       0       0       0   \n","3      4       0       0       0       0       0       0       0       0   \n","4      0       0       0       0       0       0       0       0       0   \n","\n","   pixel8  ...  pixel774  pixel775  pixel776  pixel777  pixel778  pixel779  \\\n","0       0  ...         0         0         0         0         0         0   \n","1       0  ...         0         0         0         0         0         0   \n","2       0  ...         0         0         0         0         0         0   \n","3       0  ...         0         0         0         0         0         0   \n","4       0  ...         0         0         0         0         0         0   \n","\n","   pixel780  pixel781  pixel782  pixel783  \n","0         0         0         0         0  \n","1         0         0         0         0  \n","2         0         0         0         0  \n","3         0         0         0         0  \n","4         0         0         0         0  \n","\n","[5 rows x 785 columns]"]},"execution_count":7,"metadata":{},"output_type":"execute_result"}],"source":["train_df.head()"]},{"cell_type":"code","execution_count":8,"id":"2681e162","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:37.497131Z","iopub.status.busy":"2023-04-05T14:21:37.496834Z","iopub.status.idle":"2023-04-05T14:21:37.596178Z","shell.execute_reply":"2023-04-05T14:21:37.595221Z"},"papermill":{"duration":0.109813,"end_time":"2023-04-05T14:21:37.598389","exception":false,"start_time":"2023-04-05T14:21:37.488576","status":"completed"},"tags":[]},"outputs":[{"data":{"text/plain":["((42000, 784), (42000,))"]},"execution_count":8,"metadata":{},"output_type":"execute_result"}],"source":["#Obtain Features and labels. \n","X= train_df.drop('label', axis=1)\n","y = train_df['label']\n","\n","X.shape, y.shape"]},{"cell_type":"code","execution_count":9,"id":"3957c841","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:37.613487Z","iopub.status.busy":"2023-04-05T14:21:37.613155Z","iopub.status.idle":"2023-04-05T14:21:37.878951Z","shell.execute_reply":"2023-04-05T14:21:37.877859Z"},"papermill":{"duration":0.276015,"end_time":"2023-04-05T14:21:37.881275","exception":false,"start_time":"2023-04-05T14:21:37.60526","status":"completed"},"tags":[]},"outputs":[{"data":{"text/plain":["((42000, 28, 28, 1), (28000, 28, 28, 1))"]},"execution_count":9,"metadata":{},"output_type":"execute_result"}],"source":["X = np.array(X).reshape(X.shape[0], 28, 28)\n","X = np.array(X).reshape((-1, 28, 28, 1))\n","X_test = np.array(test_df).reshape((-1, 28, 28, 1))\n","X.shape, X_test.shape"]},{"cell_type":"code","execution_count":10,"id":"5c4fed20","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:37.896324Z","iopub.status.busy":"2023-04-05T14:21:37.895992Z","iopub.status.idle":"2023-04-05T14:21:38.225015Z","shell.execute_reply":"2023-04-05T14:21:38.224114Z"},"papermill":{"duration":0.339202,"end_time":"2023-04-05T14:21:38.227236","exception":false,"start_time":"2023-04-05T14:21:37.888034","status":"completed"},"tags":[]},"outputs":[{"data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 8 Axes>"]},"metadata":{"needs_background":"light"},"output_type":"display_data"}],"source":["show_img_dataset(X, y = y, firstimg=780, nrows = 2, ncols=4, numimg=8)"]},{"cell_type":"code","execution_count":11,"id":"d7bdf943","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:38.243062Z","iopub.status.busy":"2023-04-05T14:21:38.242769Z","iopub.status.idle":"2023-04-05T14:21:38.550369Z","shell.execute_reply":"2023-04-05T14:21:38.549483Z"},"papermill":{"duration":0.318023,"end_time":"2023-04-05T14:21:38.552724","exception":false,"start_time":"2023-04-05T14:21:38.234701","status":"completed"},"tags":[]},"outputs":[{"data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 8 Axes>"]},"metadata":{"needs_background":"light"},"output_type":"display_data"}],"source":["show_img_dataset(X_test, firstimg=780, nrows = 2, ncols=4, numimg=8)"]},{"cell_type":"code","execution_count":12,"id":"0a4aad8e","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:38.569187Z","iopub.status.busy":"2023-04-05T14:21:38.568891Z","iopub.status.idle":"2023-04-05T14:21:38.743797Z","shell.execute_reply":"2023-04-05T14:21:38.742843Z"},"papermill":{"duration":0.185596,"end_time":"2023-04-05T14:21:38.746121","exception":false,"start_time":"2023-04-05T14:21:38.560525","status":"completed"},"tags":[]},"outputs":[{"data":{"text/plain":["((42000, 28, 28, 1), (42000, 10))"]},"execution_count":12,"metadata":{},"output_type":"execute_result"}],"source":["#Normalizating Features and One Hot Encoding Labels. \n","X = X/255\n","X_test = X_test/255\n","\n","y = train_df['label']\n","y = tf.keras.utils.to_categorical(y, num_classes = 10)\n","X.shape, y.shape"]},{"cell_type":"markdown","id":"29ae9bb5","metadata":{"papermill":{"duration":0.007422,"end_time":"2023-04-05T14:21:38.761432","exception":false,"start_time":"2023-04-05T14:21:38.75401","status":"completed"},"tags":[]},"source":["I use a really small validation dataset, because of the similariti of the images, it's difficult to overfit the model, but just in case I'm going to check that the overfitting is under control with this small Validation Dataset. "]},{"cell_type":"code","execution_count":13,"id":"c9ee535c","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:38.777924Z","iopub.status.busy":"2023-04-05T14:21:38.777617Z","iopub.status.idle":"2023-04-05T14:21:39.156469Z","shell.execute_reply":"2023-04-05T14:21:39.155481Z"},"papermill":{"duration":0.390087,"end_time":"2023-04-05T14:21:39.158902","exception":false,"start_time":"2023-04-05T14:21:38.768815","status":"completed"},"tags":[]},"outputs":[],"source":["#obtainig from X and y training and validating data. \n","X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.15, random_state=42)"]},{"cell_type":"code","execution_count":14,"id":"bfb833ce","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:39.17735Z","iopub.status.busy":"2023-04-05T14:21:39.17559Z","iopub.status.idle":"2023-04-05T14:21:39.181933Z","shell.execute_reply":"2023-04-05T14:21:39.181015Z"},"papermill":{"duration":0.0174,"end_time":"2023-04-05T14:21:39.184057","exception":false,"start_time":"2023-04-05T14:21:39.166657","status":"completed"},"tags":[]},"outputs":[],"source":["#just to cleant a little bit the memory, but it's no t necessary. \n","del X, y, train_df, test_df"]},{"cell_type":"code","execution_count":15,"id":"415756bc","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:39.201063Z","iopub.status.busy":"2023-04-05T14:21:39.200729Z","iopub.status.idle":"2023-04-05T14:21:39.208118Z","shell.execute_reply":"2023-04-05T14:21:39.207257Z"},"papermill":{"duration":0.018219,"end_time":"2023-04-05T14:21:39.21019","exception":false,"start_time":"2023-04-05T14:21:39.191971","status":"completed"},"tags":[]},"outputs":[{"data":{"text/plain":["((35700, 28, 28, 1), (6300, 28, 28, 1), (35700, 10), (6300, 10))"]},"execution_count":15,"metadata":{},"output_type":"execute_result"}],"source":["#Check the shapes. \n","X_train.shape, X_val.shape, y_train.shape, y_val.shape"]},{"cell_type":"markdown","id":"3250edba","metadata":{"papermill":{"duration":0.007426,"end_time":"2023-04-05T14:21:39.225534","exception":false,"start_time":"2023-04-05T14:21:39.218108","status":"completed"},"tags":[]},"source":["## Data Augmentation\n","Just a little bit of DataAugmentation. \n","\n","Vertical or horizontal flips are contraproducent, but\n","a bit of shift and zoom helps with the final accuracy. "]},{"cell_type":"code","execution_count":16,"id":"fe07a1da","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:39.242346Z","iopub.status.busy":"2023-04-05T14:21:39.242026Z","iopub.status.idle":"2023-04-05T14:21:39.47742Z","shell.execute_reply":"2023-04-05T14:21:39.476395Z"},"papermill":{"duration":0.246594,"end_time":"2023-04-05T14:21:39.479959","exception":false,"start_time":"2023-04-05T14:21:39.233365","status":"completed"},"tags":[]},"outputs":[],"source":["from keras.preprocessing.image import ImageDataGenerator\n","\n","x_trainr = X_train.reshape(-1, 28, 28, 1)\n","x_valr = X_val.reshape(-1, 28, 28, 1)\n","\n","datagen = ImageDataGenerator(\n","        rotation_range=5, \n","        zoom_range = 0.1, \n","        width_shift_range=0.05, \n","        height_shift_range=0.05,    \n",")  \n","\n","X_mean = X_train.mean(axis=0)\n","datagen.fit(X_train - X_mean)\n","train_gen = datagen.flow(x_trainr, y_train, batch_size=100)\n","val_gen = datagen.flow(x_valr, y_val, batch_size=100)"]},{"cell_type":"markdown","id":"6bf2c64f","metadata":{"papermill":{"duration":0.007785,"end_time":"2023-04-05T14:21:39.495704","exception":false,"start_time":"2023-04-05T14:21:39.487919","status":"completed"},"tags":[]},"source":["# The Model & Callback functions\n","I used two callbacks. \n","\n","**ModelCheckPoint**: to save the best model and load it after the fit. Rareley the result of the last epoch is the best. In the param **monitor** you can iondicate wich value you want to watch, the default is loss, but I prefer to improve the val_accuracy. Not sure if we can get a better score in the test dataset if we get the model with the best loss value instead of the one with the best accuracy. \n","\n","**ReduceLROnPlateau**: It reduces the **learning_rate** after the **epochs** indicated when there are no improvement in the **monitor** variable indicated. We are going to train just for 30 epochs and I indicated a really short patiente of 3 **epochs**. \n","\n","Maybe it's important to mention that I replaced the Dropout layer for a SpatialDropout one. Is similar, but it affects to an entirely channel. \n","\n","**Dropout** \n","[[1, 1, 1, 1]\n","[1, 1, 1, 1]\n","[1, 1, 1, 1]]\n","\n","Transforms to: \n","\n","[[0, 1, 1, 1]\n","[1, 1, 0, 1]\n","[1, 0, 1, 1]]\n","\n","\n","**Spatial Dropout **\n","\n","[[1, 1, 1, 1]\n","[1, 1, 1, 1]\n","[1, 1, 1, 1]]\n","\n","Transforms to: \n","\n","[[1, 1, 0, 1]\n","[1, 1, 0, 1]\n","[1, 1, 0, 1]]\n"]},{"cell_type":"code","execution_count":17,"id":"695cffac","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:39.513692Z","iopub.status.busy":"2023-04-05T14:21:39.513248Z","iopub.status.idle":"2023-04-05T14:21:39.519969Z","shell.execute_reply":"2023-04-05T14:21:39.519063Z"},"papermill":{"duration":0.018877,"end_time":"2023-04-05T14:21:39.522305","exception":false,"start_time":"2023-04-05T14:21:39.503428","status":"completed"},"tags":[]},"outputs":[],"source":["#This callback function reduces de learning rate in epochs > 20, \n","#but I didn't use in the final version. \n","def scheduler(epoch, lr):\n","    if epoch < 20:\n","        return lr\n","    else:\n","        return lr * tf.math.exp(-0.1)  \n","\n","reduce_last_lr = tf.keras.callbacks.LearningRateScheduler(scheduler)"]},{"cell_type":"code","execution_count":18,"id":"2951f5ae","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:39.538831Z","iopub.status.busy":"2023-04-05T14:21:39.538549Z","iopub.status.idle":"2023-04-05T14:21:42.17701Z","shell.execute_reply":"2023-04-05T14:21:42.1751Z"},"papermill":{"duration":2.649416,"end_time":"2023-04-05T14:21:42.179263","exception":false,"start_time":"2023-04-05T14:21:39.529847","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["Model: \"sequential\"\n","_________________________________________________________________\n","Layer (type)                 Output Shape              Param #   \n","=================================================================\n","conv2d (Conv2D)              (None, 26, 26, 32)        320       \n","_________________________________________________________________\n","max_pooling2d (MaxPooling2D) (None, 25, 25, 32)        0         \n","_________________________________________________________________\n","conv2d_1 (Conv2D)            (None, 23, 23, 32)        9248      \n","_________________________________________________________________\n","batch_normalization (BatchNo (None, 23, 23, 32)        128       \n","_________________________________________________________________\n","conv2d_2 (Conv2D)            (None, 21, 21, 32)        9248      \n","_________________________________________________________________\n","batch_normalization_1 (Batch (None, 21, 21, 32)        128       \n","_________________________________________________________________\n","spatial_dropout2d (SpatialDr (None, 21, 21, 32)        0         \n","_________________________________________________________________\n","conv2d_3 (Conv2D)            (None, 19, 19, 64)        18496     \n","_________________________________________________________________\n","max_pooling2d_1 (MaxPooling2 (None, 18, 18, 64)        0         \n","_________________________________________________________________\n","conv2d_4 (Conv2D)            (None, 16, 16, 64)        36928     \n","_________________________________________________________________\n","batch_normalization_2 (Batch (None, 16, 16, 64)        256       \n","_________________________________________________________________\n","conv2d_5 (Conv2D)            (None, 14, 14, 64)        36928     \n","_________________________________________________________________\n","batch_normalization_3 (Batch (None, 14, 14, 64)        256       \n","_________________________________________________________________\n","dropout (Dropout)            (None, 14, 14, 64)        0         \n","_________________________________________________________________\n","flatten (Flatten)            (None, 12544)             0         \n","_________________________________________________________________\n","dense (Dense)                (None, 128)               1605760   \n","_________________________________________________________________\n","lambda (Lambda)              (None, 128)               0         \n","_________________________________________________________________\n","batch_normalization_4 (Batch (None, 128)               512       \n","_________________________________________________________________\n","dropout_1 (Dropout)          (None, 128)               0         \n","_________________________________________________________________\n","dense_1 (Dense)              (None, 128)               16512     \n","_________________________________________________________________\n","lambda_1 (Lambda)            (None, 128)               0         \n","_________________________________________________________________\n","dense_2 (Dense)              (None, 10)                1290      \n","=================================================================\n","Total params: 1,736,010\n","Trainable params: 1,735,370\n","Non-trainable params: 640\n","_________________________________________________________________\n"]}],"source":["#just clean the sessio, recomendable if we execute some times the model. \n","keras.backend.clear_session()\n","\n","#This callback saves the best model based in val_accuracy\n","cpDA4 = tf.keras.callbacks.ModelCheckpoint('modelDA.h5', \n","                                           mode='max', monitor='val_accuracy', \n","                                           verbose=1, \n","                                           save_best_only=True)\n","\n","#This callback will reduce the learning rate after 3 epochs with no improvement in \n","#the val_accuracy. \n","reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(monitor='val_accuracy', factor=0.1,\n","                              patience=3, verbose=1, mode='auto')\n","#This callback reduces the lr in the last epochs, but i'm not using it by the moment. \n","reduce_last_lr = tf.keras.callbacks.LearningRateScheduler(scheduler)\n","\n","modelDA4 = tf.keras.models.Sequential([\n","    tf.keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(28, 28, 1)),\n","    tf.keras.layers.MaxPooling2D((2,2), strides=(1, 1), padding='valid'),\n","    tf.keras.layers.Conv2D(32, (3,3), activation='relu'),\n","    tf.keras.layers.BatchNormalization(),\n","    tf.keras.layers.Conv2D(32, (3,3), activation='relu'),\n","    tf.keras.layers.BatchNormalization(),\n","    tf.keras.layers.SpatialDropout2D(0.2),\n","\n","    tf.keras.layers.Conv2D(64, (3,3), activation='relu'),\n","    tf.keras.layers.MaxPooling2D((2,2), strides=(1, 1), padding='valid'),\n","    tf.keras.layers.Conv2D(64, (3,3), activation='relu'),\n","    tf.keras.layers.BatchNormalization(),\n","    tf.keras.layers.Conv2D(64, (3,3), activation='relu'),\n","    tf.keras.layers.BatchNormalization(),\n","    tf.keras.layers.Dropout(0.2),\n","\n","    tf.keras.layers.Flatten(),\n","    tf.keras.layers.Dense(128),\n","    tf.keras.layers.Lambda(lambda x: tf.abs(x)),\n","    tf.keras.layers.BatchNormalization(),\n","    tf.keras.layers.Dropout(0.2),  \n","    tf.keras.layers.Dense(128),\n","    tf.keras.layers.Lambda(lambda x: tf.abs(x)),\n","    tf.keras.layers.Dense(10, activation='softmax')])\n","modelDA4.summary()"]},{"cell_type":"code","execution_count":19,"id":"13d805b7","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:21:42.196995Z","iopub.status.busy":"2023-04-05T14:21:42.196706Z","iopub.status.idle":"2023-04-05T14:27:24.794021Z","shell.execute_reply":"2023-04-05T14:27:24.792015Z"},"papermill":{"duration":342.608527,"end_time":"2023-04-05T14:27:24.796185","exception":false,"start_time":"2023-04-05T14:21:42.187658","status":"completed"},"scrolled":true,"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["Epoch 1/30\n","357/357 [==============================] - 19s 32ms/step - loss: 0.4293 - accuracy: 0.8750 - val_loss: 1.5978 - val_accuracy: 0.5338\n","\n","Epoch 00001: val_accuracy improved from -inf to 0.53381, saving model to modelDA.h5\n","Epoch 2/30\n","357/357 [==============================] - 10s 29ms/step - loss: 0.0982 - accuracy: 0.9722 - val_loss: 0.0706 - val_accuracy: 0.9789\n","\n","Epoch 00002: val_accuracy improved from 0.53381 to 0.97889, saving model to modelDA.h5\n","Epoch 3/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0712 - accuracy: 0.9796 - val_loss: 0.0536 - val_accuracy: 0.9837\n","\n","Epoch 00003: val_accuracy improved from 0.97889 to 0.98365, saving model to modelDA.h5\n","Epoch 4/30\n","357/357 [==============================] - 11s 31ms/step - loss: 0.0604 - accuracy: 0.9823 - val_loss: 0.0815 - val_accuracy: 0.9765\n","\n","Epoch 00004: val_accuracy did not improve from 0.98365\n","Epoch 5/30\n","357/357 [==============================] - 11s 29ms/step - loss: 0.0537 - accuracy: 0.9846 - val_loss: 0.0809 - val_accuracy: 0.9767\n","\n","Epoch 00005: val_accuracy did not improve from 0.98365\n","Epoch 6/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0506 - accuracy: 0.9847 - val_loss: 0.0540 - val_accuracy: 0.9841\n","\n","Epoch 00006: val_accuracy improved from 0.98365 to 0.98413, saving model to modelDA.h5\n","Epoch 7/30\n","357/357 [==============================] - 11s 31ms/step - loss: 0.0439 - accuracy: 0.9864 - val_loss: 0.0358 - val_accuracy: 0.9883\n","\n","Epoch 00007: val_accuracy improved from 0.98413 to 0.98825, saving model to modelDA.h5\n","Epoch 8/30\n","357/357 [==============================] - 11s 29ms/step - loss: 0.0412 - accuracy: 0.9874 - val_loss: 0.0376 - val_accuracy: 0.9883\n","\n","Epoch 00008: val_accuracy did not improve from 0.98825\n","Epoch 9/30\n","357/357 [==============================] - 11s 31ms/step - loss: 0.0377 - accuracy: 0.9886 - val_loss: 0.0412 - val_accuracy: 0.9876\n","\n","Epoch 00009: val_accuracy did not improve from 0.98825\n","Epoch 10/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0390 - accuracy: 0.9884 - val_loss: 0.0433 - val_accuracy: 0.9884\n","\n","Epoch 00010: val_accuracy improved from 0.98825 to 0.98841, saving model to modelDA.h5\n","Epoch 11/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0339 - accuracy: 0.9893 - val_loss: 0.0415 - val_accuracy: 0.9889\n","\n","Epoch 00011: val_accuracy improved from 0.98841 to 0.98889, saving model to modelDA.h5\n","Epoch 12/30\n","357/357 [==============================] - 11s 31ms/step - loss: 0.0333 - accuracy: 0.9899 - val_loss: 0.0488 - val_accuracy: 0.9863\n","\n","Epoch 00012: val_accuracy did not improve from 0.98889\n","Epoch 13/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0303 - accuracy: 0.9906 - val_loss: 0.0448 - val_accuracy: 0.9879\n","\n","Epoch 00013: val_accuracy did not improve from 0.98889\n","Epoch 14/30\n","357/357 [==============================] - 10s 29ms/step - loss: 0.0327 - accuracy: 0.9901 - val_loss: 0.0469 - val_accuracy: 0.9870\n","\n","Epoch 00014: val_accuracy did not improve from 0.98889\n","\n","Epoch 00014: ReduceLROnPlateau reducing learning rate to 0.00010000000474974513.\n","Epoch 15/30\n","357/357 [==============================] - 11s 31ms/step - loss: 0.0204 - accuracy: 0.9939 - val_loss: 0.0316 - val_accuracy: 0.9914\n","\n","Epoch 00015: val_accuracy improved from 0.98889 to 0.99143, saving model to modelDA.h5\n","Epoch 16/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0162 - accuracy: 0.9949 - val_loss: 0.0211 - val_accuracy: 0.9927\n","\n","Epoch 00016: val_accuracy improved from 0.99143 to 0.99270, saving model to modelDA.h5\n","Epoch 17/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0124 - accuracy: 0.9964 - val_loss: 0.0221 - val_accuracy: 0.9925\n","\n","Epoch 00017: val_accuracy did not improve from 0.99270\n","Epoch 18/30\n","357/357 [==============================] - 11s 31ms/step - loss: 0.0126 - accuracy: 0.9959 - val_loss: 0.0208 - val_accuracy: 0.9935\n","\n","Epoch 00018: val_accuracy improved from 0.99270 to 0.99349, saving model to modelDA.h5\n","Epoch 19/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0118 - accuracy: 0.9961 - val_loss: 0.0242 - val_accuracy: 0.9922\n","\n","Epoch 00019: val_accuracy did not improve from 0.99349\n","Epoch 20/30\n","357/357 [==============================] - 10s 29ms/step - loss: 0.0121 - accuracy: 0.9962 - val_loss: 0.0249 - val_accuracy: 0.9927\n","\n","Epoch 00020: val_accuracy did not improve from 0.99349\n","Epoch 21/30\n","357/357 [==============================] - 11s 32ms/step - loss: 0.0116 - accuracy: 0.9962 - val_loss: 0.0237 - val_accuracy: 0.9940\n","\n","Epoch 00021: val_accuracy improved from 0.99349 to 0.99397, saving model to modelDA.h5\n","Epoch 22/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0110 - accuracy: 0.9966 - val_loss: 0.0200 - val_accuracy: 0.9940\n","\n","Epoch 00022: val_accuracy did not improve from 0.99397\n","Epoch 23/30\n","357/357 [==============================] - 10s 29ms/step - loss: 0.0105 - accuracy: 0.9966 - val_loss: 0.0196 - val_accuracy: 0.9937\n","\n","Epoch 00023: val_accuracy did not improve from 0.99397\n","Epoch 24/30\n","357/357 [==============================] - 11s 31ms/step - loss: 0.0090 - accuracy: 0.9973 - val_loss: 0.0218 - val_accuracy: 0.9933\n","\n","Epoch 00024: val_accuracy did not improve from 0.99397\n","\n","Epoch 00024: ReduceLROnPlateau reducing learning rate to 1.0000000474974514e-05.\n","Epoch 25/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0084 - accuracy: 0.9977 - val_loss: 0.0191 - val_accuracy: 0.9935\n","\n","Epoch 00025: val_accuracy did not improve from 0.99397\n","Epoch 26/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0094 - accuracy: 0.9971 - val_loss: 0.0202 - val_accuracy: 0.9937\n","\n","Epoch 00026: val_accuracy did not improve from 0.99397\n","Epoch 27/30\n","357/357 [==============================] - 11s 32ms/step - loss: 0.0086 - accuracy: 0.9972 - val_loss: 0.0184 - val_accuracy: 0.9938\n","\n","Epoch 00027: val_accuracy did not improve from 0.99397\n","\n","Epoch 00027: ReduceLROnPlateau reducing learning rate to 1.0000000656873453e-06.\n","Epoch 28/30\n","357/357 [==============================] - 11s 29ms/step - loss: 0.0082 - accuracy: 0.9975 - val_loss: 0.0216 - val_accuracy: 0.9932\n","\n","Epoch 00028: val_accuracy did not improve from 0.99397\n","Epoch 29/30\n","357/357 [==============================] - 11s 30ms/step - loss: 0.0081 - accuracy: 0.9975 - val_loss: 0.0198 - val_accuracy: 0.9938\n","\n","Epoch 00029: val_accuracy did not improve from 0.99397\n","Epoch 30/30\n","357/357 [==============================] - 11s 31ms/step - loss: 0.0082 - accuracy: 0.9974 - val_loss: 0.0202 - val_accuracy: 0.9941\n","\n","Epoch 00030: val_accuracy improved from 0.99397 to 0.99413, saving model to modelDA.h5\n"]}],"source":["modelDA4.compile(optimizer='adam', \n","                  loss='categorical_crossentropy', \n","                  metrics=['accuracy']) \n","\n","\n","historyDA4 = modelDA4.fit(train_gen, \n","                        steps_per_epoch=len(train_gen),\n","                        validation_data=val_gen,\n","                        validation_steps=len(val_gen), \n","                        epochs=30, \n","                        verbose=1, \n","                        callbacks=[cpDA4, reduce_lr]\n","                        )"]},{"cell_type":"markdown","id":"39a63358","metadata":{"papermill":{"duration":0.248327,"end_time":"2023-04-05T14:27:25.294425","exception":false,"start_time":"2023-04-05T14:27:25.046098","status":"completed"},"tags":[]},"source":["# Evaluate results"]},{"cell_type":"code","execution_count":20,"id":"09e266c0","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:27:25.842841Z","iopub.status.busy":"2023-04-05T14:27:25.84248Z","iopub.status.idle":"2023-04-05T14:27:26.512263Z","shell.execute_reply":"2023-04-05T14:27:26.51126Z"},"papermill":{"duration":0.921847,"end_time":"2023-04-05T14:27:26.514683","exception":false,"start_time":"2023-04-05T14:27:25.592836","status":"completed"},"tags":[]},"outputs":[{"data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"},"output_type":"display_data"},{"data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"needs_background":"light"},"output_type":"display_data"}],"source":["plot_loss_acc(historyDA4, )\n","modelDA_loaded = load_model('modelDA.h5')"]},{"cell_type":"code","execution_count":21,"id":"ad228354","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:27:27.015864Z","iopub.status.busy":"2023-04-05T14:27:27.015138Z","iopub.status.idle":"2023-04-05T14:27:28.070224Z","shell.execute_reply":"2023-04-05T14:27:28.069214Z"},"papermill":{"duration":1.310041,"end_time":"2023-04-05T14:27:28.072453","exception":false,"start_time":"2023-04-05T14:27:26.762412","status":"completed"},"tags":[]},"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>Predictions</th>\n","      <th>Actuals</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>8</td>\n","      <td>8</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>1</td>\n","      <td>1</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>9</td>\n","      <td>9</td>\n","    </tr>\n","    <tr>\n","      <th>3</th>\n","      <td>9</td>\n","      <td>9</td>\n","    </tr>\n","    <tr>\n","      <th>4</th>\n","      <td>8</td>\n","      <td>8</td>\n","    </tr>\n","    <tr>\n","      <th>...</th>\n","      <td>...</td>\n","      <td>...</td>\n","    </tr>\n","    <tr>\n","      <th>6295</th>\n","      <td>4</td>\n","      <td>4</td>\n","    </tr>\n","    <tr>\n","      <th>6296</th>\n","      <td>1</td>\n","      <td>1</td>\n","    </tr>\n","    <tr>\n","      <th>6297</th>\n","      <td>6</td>\n","      <td>6</td>\n","    </tr>\n","    <tr>\n","      <th>6298</th>\n","      <td>5</td>\n","      <td>5</td>\n","    </tr>\n","    <tr>\n","      <th>6299</th>\n","      <td>0</td>\n","      <td>0</td>\n","    </tr>\n","  </tbody>\n","</table>\n","<p>6300 rows × 2 columns</p>\n","</div>"],"text/plain":["      Predictions  Actuals\n","0               8        8\n","1               1        1\n","2               9        9\n","3               9        9\n","4               8        8\n","...           ...      ...\n","6295            4        4\n","6296            1        1\n","6297            6        6\n","6298            5        5\n","6299            0        0\n","\n","[6300 rows x 2 columns]"]},"execution_count":21,"metadata":{},"output_type":"execute_result"}],"source":["prediction, results = get_predictions(modelDA_loaded, X_val, y_val)\n","results"]},{"cell_type":"code","execution_count":22,"id":"f5e933c2","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:27:28.580578Z","iopub.status.busy":"2023-04-05T14:27:28.579524Z","iopub.status.idle":"2023-04-05T14:27:30.340846Z","shell.execute_reply":"2023-04-05T14:27:30.339379Z"},"papermill":{"duration":2.018435,"end_time":"2023-04-05T14:27:30.342949","exception":false,"start_time":"2023-04-05T14:27:28.324514","status":"completed"},"tags":[]},"outputs":[{"name":"stdout","output_type":"stream","text":["197/197 [==============================] - 1s 3ms/step - loss: 0.0175 - accuracy: 0.9941\n","197/197 [==============================] - 1s 3ms/step - loss: 0.0175 - accuracy: 0.9941\n","Validation loss:  0.01746087148785591\n","Validation accuracy:  0.9941269755363464\n","Validation loss loaded:  0.01746087148785591\n","Validation accuracy loaded:  0.9941269755363464\n"]}],"source":["validation_loss, validation_accuracy = modelDA4.evaluate(X_val, y_val)\n","validation_loss_loaded, validation_accuracy_loaded = modelDA_loaded.evaluate(X_val, y_val)\n","print('Validation loss: ', validation_loss)\n","print('Validation accuracy: ', validation_accuracy)\n","\n","print('Validation loss loaded: ', validation_loss_loaded)\n","print('Validation accuracy loaded: ', validation_accuracy_loaded)"]},{"cell_type":"code","execution_count":23,"id":"65af9742","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:27:30.896956Z","iopub.status.busy":"2023-04-05T14:27:30.896593Z","iopub.status.idle":"2023-04-05T14:27:33.808152Z","shell.execute_reply":"2023-04-05T14:27:33.807099Z"},"papermill":{"duration":3.166174,"end_time":"2023-04-05T14:27:33.810431","exception":false,"start_time":"2023-04-05T14:27:30.644257","status":"completed"},"tags":[]},"outputs":[{"data":{"text/plain":["((28000,), array([2, 0, 9, ..., 3, 9, 2]))"]},"execution_count":23,"metadata":{},"output_type":"execute_result"}],"source":["\n","y_pred=modelDA_loaded.predict(X_test)\n","results = np.argmax(y_pred, axis=1)\n","results.shape, results"]},{"cell_type":"code","execution_count":24,"id":"97d4f1d2","metadata":{"execution":{"iopub.execute_input":"2023-04-05T14:27:34.315883Z","iopub.status.busy":"2023-04-05T14:27:34.315523Z","iopub.status.idle":"2023-04-05T14:27:34.347999Z","shell.execute_reply":"2023-04-05T14:27:34.347049Z"},"papermill":{"duration":0.286478,"end_time":"2023-04-05T14:27:34.350351","exception":false,"start_time":"2023-04-05T14:27:34.063873","status":"completed"},"tags":[]},"outputs":[],"source":["image_id = pd.Series(range(1,28001),name='ImageId')\n","y_preds = pd.Series(results,name = 'Label')\n","pred = pd.concat([image_id,y_preds],axis=1)\n","pred.to_csv('submission.csv',index=False)"]}],"metadata":{"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":"3.7.12"},"papermill":{"default_parameters":{},"duration":379.4868,"end_time":"2023-04-05T14:27:38.310746","environment_variables":{},"exception":null,"input_path":"__notebook__.ipynb","output_path":"__notebook__.ipynb","parameters":{},"start_time":"2023-04-05T14:21:18.823946","version":"2.3.4"}},"nbformat":4,"nbformat_minor":5}