FocalLoss_Ex.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"FocalLoss_Trial.ipynb","provenance":[],"collapsed_sections":[],"authorship_tag":"ABX9TyMhq2Re7fmFFpQA1OkBdIpw"},"kernelspec":{"name":"python3","display_name":"Python 3"}},"cells":[{"cell_type":"code","metadata":{"id":"1c7GRJpVG9ga","executionInfo":{"status":"ok","timestamp":1612796776560,"user_tz":-60,"elapsed":4161,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}}},"source":["import tensorflow as tf\n","\n","import numpy as np \n","import pandas as pd \n","import matplotlib.pyplot as plt \n","import seaborn as sns \n","\n","\n","#building all kinds of evaluating parameters\n","from sklearn.model_selection import train_test_split\n","from sklearn.metrics import classification_report, accuracy_score  \n","from sklearn.metrics import precision_score, recall_score \n","from sklearn.metrics import f1_score, matthews_corrcoef \n","from sklearn.metrics import confusion_matrix \n","\n","from sklearn.preprocessing import RobustScaler"],"execution_count":1,"outputs":[]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"RYzdxqtPZ-g0","executionInfo":{"status":"ok","timestamp":1612391668918,"user_tz":-60,"elapsed":23353,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"5c184f15-d3f7-46c0-e36b-27d12659320e"},"source":["from google.colab import drive\n","drive.mount('/content/drive/')"],"execution_count":null,"outputs":[{"output_type":"stream","text":["Mounted at /content/drive/\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/","height":462},"id":"RXMJV0RN_TXa","executionInfo":{"status":"ok","timestamp":1612646871513,"user_tz":-60,"elapsed":2629,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"79e6eb38-f73c-470d-9011-e29c3090b1da"},"source":["p = np.linspace(0.01, 1.0, 50)\n","# print (p)\n","logp = -np.log(p)\n","ones = np.ones(len(p))\n","def fterm(gamma):\n","  return (ones - p)**(gamma)\n","\n","fl2 = fterm(2) * logp\n","\n","fl0 = fterm(0) * logp\n","fl5 = fterm(5) * logp\n","fl0d5 = fterm(0.5) * logp\n","\n","fig=plt.figure(figsize=(9, 7))\n","plt.plot(p, logp, 'b-', label='CE Loss')\n","\n","plt.plot(p, fl0d5, linestyle='-.', color='magenta', label=r'Focal Loss: $\\gamma=0.5$')\n","plt.plot(p, fl2, 'r.', label=r'Focal Loss: $\\gamma=2$')\n","plt.plot(p, fl5, linestyle='--', color='orange', label=r'Focal Loss: $\\gamma=5$')\n","# plt.text(0.1, 4, )\n","plt.legend(fontsize=12)\n","plt.xlabel('Probability of Ground Truth Class', fontsize=12)\n","plt.ylabel('Loss', fontsize=12)\n","\n","# print (fterm(2))\n","# plt.savefig('/content/drive/My Drive/Colab Notebooks/Focal_Loss.png', dpi=200)"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["Text(0, 0.5, 'Loss')"]},"metadata":{"tags":[]},"execution_count":2},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 648x504 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"code","metadata":{"id":"8_KR5wOdHDSF","colab":{"base_uri":"https://localhost:8080/","height":163},"executionInfo":{"status":"ok","timestamp":1610886325887,"user_tz":-60,"elapsed":2836,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"e488c965-e158-4891-b6ff-36505da1cedc"},"source":["credit_df = pd.read_csv('/content/drive/My Drive/Colab Notebooks/creditcard.csv', sep=',')\n","credit_df.head(3)"],"execution_count":null,"outputs":[{"output_type":"execute_result","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>Time</th>\n","      <th>V1</th>\n","      <th>V2</th>\n","      <th>V3</th>\n","      <th>V4</th>\n","      <th>V5</th>\n","      <th>V6</th>\n","      <th>V7</th>\n","      <th>V8</th>\n","      <th>V9</th>\n","      <th>V10</th>\n","      <th>V11</th>\n","      <th>V12</th>\n","      <th>V13</th>\n","      <th>V14</th>\n","      <th>V15</th>\n","      <th>V16</th>\n","      <th>V17</th>\n","      <th>V18</th>\n","      <th>V19</th>\n","      <th>V20</th>\n","      <th>V21</th>\n","      <th>V22</th>\n","      <th>V23</th>\n","      <th>V24</th>\n","      <th>V25</th>\n","      <th>V26</th>\n","      <th>V27</th>\n","      <th>V28</th>\n","      <th>Amount</th>\n","      <th>Class</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>0.0</td>\n","      <td>-1.359807</td>\n","      <td>-0.072781</td>\n","      <td>2.536347</td>\n","      <td>1.378155</td>\n","      <td>-0.338321</td>\n","      <td>0.462388</td>\n","      <td>0.239599</td>\n","      <td>0.098698</td>\n","      <td>0.363787</td>\n","      <td>0.090794</td>\n","      <td>-0.551600</td>\n","      <td>-0.617801</td>\n","      <td>-0.991390</td>\n","      <td>-0.311169</td>\n","      <td>1.468177</td>\n","      <td>-0.470401</td>\n","      <td>0.207971</td>\n","      <td>0.025791</td>\n","      <td>0.403993</td>\n","      <td>0.251412</td>\n","      <td>-0.018307</td>\n","      <td>0.277838</td>\n","      <td>-0.110474</td>\n","      <td>0.066928</td>\n","      <td>0.128539</td>\n","      <td>-0.189115</td>\n","      <td>0.133558</td>\n","      <td>-0.021053</td>\n","      <td>149.62</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>0.0</td>\n","      <td>1.191857</td>\n","      <td>0.266151</td>\n","      <td>0.166480</td>\n","      <td>0.448154</td>\n","      <td>0.060018</td>\n","      <td>-0.082361</td>\n","      <td>-0.078803</td>\n","      <td>0.085102</td>\n","      <td>-0.255425</td>\n","      <td>-0.166974</td>\n","      <td>1.612727</td>\n","      <td>1.065235</td>\n","      <td>0.489095</td>\n","      <td>-0.143772</td>\n","      <td>0.635558</td>\n","      <td>0.463917</td>\n","      <td>-0.114805</td>\n","      <td>-0.183361</td>\n","      <td>-0.145783</td>\n","      <td>-0.069083</td>\n","      <td>-0.225775</td>\n","      <td>-0.638672</td>\n","      <td>0.101288</td>\n","      <td>-0.339846</td>\n","      <td>0.167170</td>\n","      <td>0.125895</td>\n","      <td>-0.008983</td>\n","      <td>0.014724</td>\n","      <td>2.69</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>1.0</td>\n","      <td>-1.358354</td>\n","      <td>-1.340163</td>\n","      <td>1.773209</td>\n","      <td>0.379780</td>\n","      <td>-0.503198</td>\n","      <td>1.800499</td>\n","      <td>0.791461</td>\n","      <td>0.247676</td>\n","      <td>-1.514654</td>\n","      <td>0.207643</td>\n","      <td>0.624501</td>\n","      <td>0.066084</td>\n","      <td>0.717293</td>\n","      <td>-0.165946</td>\n","      <td>2.345865</td>\n","      <td>-2.890083</td>\n","      <td>1.109969</td>\n","      <td>-0.121359</td>\n","      <td>-2.261857</td>\n","      <td>0.524980</td>\n","      <td>0.247998</td>\n","      <td>0.771679</td>\n","      <td>0.909412</td>\n","      <td>-0.689281</td>\n","      <td>-0.327642</td>\n","      <td>-0.139097</td>\n","      <td>-0.055353</td>\n","      <td>-0.059752</td>\n","      <td>378.66</td>\n","      <td>0</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["   Time        V1        V2        V3  ...       V27       V28  Amount  Class\n","0   0.0 -1.359807 -0.072781  2.536347  ...  0.133558 -0.021053  149.62      0\n","1   0.0  1.191857  0.266151  0.166480  ... -0.008983  0.014724    2.69      0\n","2   1.0 -1.358354 -1.340163  1.773209  ... -0.055353 -0.059752  378.66      0\n","\n","[3 rows x 31 columns]"]},"metadata":{"tags":[]},"execution_count":2}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"TABPHO8CasK9","executionInfo":{"status":"ok","timestamp":1610886334837,"user_tz":-60,"elapsed":672,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"c55d79fb-b7fa-41ef-8b24-c52147f8c1f4"},"source":["print ('dataframe shape: ', credit_df.shape)"],"execution_count":null,"outputs":[{"output_type":"stream","text":["dataframe shape:  (284807, 31)\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/","height":282},"id":"dFmPixlTazLo","executionInfo":{"status":"ok","timestamp":1610886097479,"user_tz":-60,"elapsed":659,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"39f3c08c-1fc5-4b4c-8090-c1d791020baa"},"source":["# check the label distribution \n","\n","sns.countplot(x='Class', data=credit_df, palette='hls')\n","plt.xlabel('Label', fontsize=12)\n","plt.ylabel('Counts', fontsize=12)\n","plt.show()"],"execution_count":null,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"n6TODc5lcDTE"},"source":["we see here extremely imbalanced data in action here. Fradulent cases are negligible and that's why it gets difficult to identify the fraud cases ... "]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"MJTgZC9Yb_Lh","executionInfo":{"status":"ok","timestamp":1610886119587,"user_tz":-60,"elapsed":766,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"118b21c7-f637-474b-f880-d310ae1f163b"},"source":["print ('Cases without frauds: ', len(credit_df[credit_df['Class']==0]))\n","print ('Cases with frauds: ', len(credit_df[credit_df['Class']==1]))"],"execution_count":null,"outputs":[{"output_type":"stream","text":["Cases without frauds:  284315\n","Cases with frauds:  492\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"WIz5vnwTcdHD","executionInfo":{"status":"ok","timestamp":1609183698249,"user_tz":-60,"elapsed":643,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"6874701d-bb64-4493-9b4f-c08a371f8b71"},"source":["print ('check for null values in each column: ', credit_df.isna().sum())"],"execution_count":null,"outputs":[{"output_type":"stream","text":["check for null values in each column:  Time      0\n","V1        0\n","V2        0\n","V3        0\n","V4        0\n","V5        0\n","V6        0\n","V7        0\n","V8        0\n","V9        0\n","V10       0\n","V11       0\n","V12       0\n","V13       0\n","V14       0\n","V15       0\n","V16       0\n","V17       0\n","V18       0\n","V19       0\n","V20       0\n","V21       0\n","V22       0\n","V23       0\n","V24       0\n","V25       0\n","V26       0\n","V27       0\n","V28       0\n","Amount    0\n","Class     0\n","dtype: int64\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"YTX3htY40rAt"},"source":["No null values in the data-frame!! "]},{"cell_type":"markdown","metadata":{"id":"X3_aRMK57Pyf"},"source":["If you see values in the respective columns, _Amount_ and _Time_ are not scaled but other columns are scaled. So to include Amount and Time as features we need to scale them. \n","Here we can use StandardScaler/RobustScaler to scale these features.   "]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/","height":163},"id":"LG9Jwx4oFCaP","executionInfo":{"status":"ok","timestamp":1610886376514,"user_tz":-60,"elapsed":697,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"bf28cab2-9bf5-47a0-9644-64be029e9115"},"source":["## Scale Amount and Time \n","\n","robust_scaler = RobustScaler()\n","\n","credit_df[['Amount', 'Time']] = robust_scaler.fit_transform(credit_df[['Amount', 'Time']])\n","\n","credit_df.head(3)"],"execution_count":null,"outputs":[{"output_type":"execute_result","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>Time</th>\n","      <th>V1</th>\n","      <th>V2</th>\n","      <th>V3</th>\n","      <th>V4</th>\n","      <th>V5</th>\n","      <th>V6</th>\n","      <th>V7</th>\n","      <th>V8</th>\n","      <th>V9</th>\n","      <th>V10</th>\n","      <th>V11</th>\n","      <th>V12</th>\n","      <th>V13</th>\n","      <th>V14</th>\n","      <th>V15</th>\n","      <th>V16</th>\n","      <th>V17</th>\n","      <th>V18</th>\n","      <th>V19</th>\n","      <th>V20</th>\n","      <th>V21</th>\n","      <th>V22</th>\n","      <th>V23</th>\n","      <th>V24</th>\n","      <th>V25</th>\n","      <th>V26</th>\n","      <th>V27</th>\n","      <th>V28</th>\n","      <th>Amount</th>\n","      <th>Class</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0</th>\n","      <td>-0.994983</td>\n","      <td>-1.359807</td>\n","      <td>-0.072781</td>\n","      <td>2.536347</td>\n","      <td>1.378155</td>\n","      <td>-0.338321</td>\n","      <td>0.462388</td>\n","      <td>0.239599</td>\n","      <td>0.098698</td>\n","      <td>0.363787</td>\n","      <td>0.090794</td>\n","      <td>-0.551600</td>\n","      <td>-0.617801</td>\n","      <td>-0.991390</td>\n","      <td>-0.311169</td>\n","      <td>1.468177</td>\n","      <td>-0.470401</td>\n","      <td>0.207971</td>\n","      <td>0.025791</td>\n","      <td>0.403993</td>\n","      <td>0.251412</td>\n","      <td>-0.018307</td>\n","      <td>0.277838</td>\n","      <td>-0.110474</td>\n","      <td>0.066928</td>\n","      <td>0.128539</td>\n","      <td>-0.189115</td>\n","      <td>0.133558</td>\n","      <td>-0.021053</td>\n","      <td>1.783274</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>1</th>\n","      <td>-0.994983</td>\n","      <td>1.191857</td>\n","      <td>0.266151</td>\n","      <td>0.166480</td>\n","      <td>0.448154</td>\n","      <td>0.060018</td>\n","      <td>-0.082361</td>\n","      <td>-0.078803</td>\n","      <td>0.085102</td>\n","      <td>-0.255425</td>\n","      <td>-0.166974</td>\n","      <td>1.612727</td>\n","      <td>1.065235</td>\n","      <td>0.489095</td>\n","      <td>-0.143772</td>\n","      <td>0.635558</td>\n","      <td>0.463917</td>\n","      <td>-0.114805</td>\n","      <td>-0.183361</td>\n","      <td>-0.145783</td>\n","      <td>-0.069083</td>\n","      <td>-0.225775</td>\n","      <td>-0.638672</td>\n","      <td>0.101288</td>\n","      <td>-0.339846</td>\n","      <td>0.167170</td>\n","      <td>0.125895</td>\n","      <td>-0.008983</td>\n","      <td>0.014724</td>\n","      <td>-0.269825</td>\n","      <td>0</td>\n","    </tr>\n","    <tr>\n","      <th>2</th>\n","      <td>-0.994972</td>\n","      <td>-1.358354</td>\n","      <td>-1.340163</td>\n","      <td>1.773209</td>\n","      <td>0.379780</td>\n","      <td>-0.503198</td>\n","      <td>1.800499</td>\n","      <td>0.791461</td>\n","      <td>0.247676</td>\n","      <td>-1.514654</td>\n","      <td>0.207643</td>\n","      <td>0.624501</td>\n","      <td>0.066084</td>\n","      <td>0.717293</td>\n","      <td>-0.165946</td>\n","      <td>2.345865</td>\n","      <td>-2.890083</td>\n","      <td>1.109969</td>\n","      <td>-0.121359</td>\n","      <td>-2.261857</td>\n","      <td>0.524980</td>\n","      <td>0.247998</td>\n","      <td>0.771679</td>\n","      <td>0.909412</td>\n","      <td>-0.689281</td>\n","      <td>-0.327642</td>\n","      <td>-0.139097</td>\n","      <td>-0.055353</td>\n","      <td>-0.059752</td>\n","      <td>4.983721</td>\n","      <td>0</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["       Time        V1        V2        V3  ...       V27       V28    Amount  Class\n","0 -0.994983 -1.359807 -0.072781  2.536347  ...  0.133558 -0.021053  1.783274      0\n","1 -0.994983  1.191857  0.266151  0.166480  ... -0.008983  0.014724 -0.269825      0\n","2 -0.994972 -1.358354 -1.340163  1.773209  ... -0.055353 -0.059752  4.983721      0\n","\n","[3 rows x 31 columns]"]},"metadata":{"tags":[]},"execution_count":5}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"DippltKM0po3","executionInfo":{"status":"ok","timestamp":1610886381134,"user_tz":-60,"elapsed":802,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"0092564e-c612-44c7-d328-2f29e058eeeb"},"source":["X_labels = credit_df.drop(['Class'], axis=1)\n","y_labels = credit_df['Class']\n","\n","print ('x shape and type: ', X_labels.shape, type(X_labels))\n","print ('y shape: ', y_labels.shape)\n","\n","X_labels = X_labels.to_numpy(dtype=np.float64)\n","y_labels = y_labels.to_numpy(dtype=np.float64)\n","\n","print ('x shape and type: ', X_labels.shape, type(X_labels))\n","\n","print ('check few yvals: ', y_labels[0:3])\n","\n","y_lab_cat = tf.keras.utils.to_categorical(y_labels, num_classes=2, dtype='float32')\n","\n","print ('categorical y_lab: ', y_lab_cat[0:3])"],"execution_count":null,"outputs":[{"output_type":"stream","text":["x shape and type:  (284807, 30) <class 'pandas.core.frame.DataFrame'>\n","y shape:  (284807,)\n","x shape and type:  (284807, 30) <class 'numpy.ndarray'>\n","check few yvals:  [0. 0. 0.]\n","categorical y_lab:  [[1. 0.]\n"," [1. 0.]\n"," [1. 0.]]\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"7oNYws9ZG2yQ","executionInfo":{"status":"ok","timestamp":1610886395441,"user_tz":-60,"elapsed":2034,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"c0d9d240-3c72-4c50-a50a-4d0e4d75eed0"},"source":["x_train, x_test, y_train, y_test = train_test_split(X_labels, y_lab_cat, test_size=0.3, \n","                                                    stratify=y_lab_cat, shuffle=True)\n","\n","\n","# stratify ensure same proportion of class labels\n","\n","print('train data shape and type: ', x_train.shape, type(x_train))\n","print ('test data shape: ', x_test.shape)\n","\n","print ('check few y vals: ', y_train[0:3])"],"execution_count":null,"outputs":[{"output_type":"stream","text":["train data shape and type:  (199364, 30) <class 'numpy.ndarray'>\n","test data shape:  (85443, 30)\n","check few y vals:  [[1. 0.]\n"," [1. 0.]\n"," [1. 0.]]\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"hbaYwFplHrBu"},"source":["from tensorflow.keras.layers import  Dense, Input, Activation, Dropout\n","from tensorflow.keras.models import Model\n","from tensorflow.keras.optimizers import Adam\n","from tensorflow.keras import activations \n","from tensorflow import keras\n","\n","from keras import backend as K"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"NGH3veI2J4gI"},"source":["def simple_model():\n","  input_data  = Input(shape=(x_train.shape[1], ))\n","  x = Dense(64)(input_data)\n","  x = Activation(activations.relu)(x)\n","  x = Dense(32)(x)\n","  x = Activation(activations.relu)(x)\n","  x = Dense(2)(x)\n","  x = Activation(activations.softmax)(x)\n","\n","  model = Model(inputs=input_data, outputs=x, name='Simple_Model')\n","\n","  return model  "],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"W-cKfDv_8rK5"},"source":["def simple_model1():\n","  input_data  = Input(shape=(x_train.shape[1], ))\n","  x = Dense(64)(input_data)\n","  x = Activation(activations.relu)(x)\n","  x = Dense(32)(x)\n","  x = Activation(activations.relu)(x)\n","  x = Dense(2)(x)\n","  x = Activation(activations.softmax)(x)\n","\n","  model = Model(inputs=input_data, outputs=x, name='Simple_Model')\n","\n","  return model  "],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"hHaj1-jmbe6y"},"source":["simple_model = simple_model()\n","simple_model.summary()"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"57TQaOcobn9j"},"source":["# simple_model.compile(optimizer=Adam(learning_rate=1e-2), loss='sparse_categorical_crossentropy', \n","#                       metrics=['acc'])\n","\n","simple_model.compile(optimizer=Adam(learning_rate=5e-3), loss='categorical_crossentropy', \n","                      metrics=['acc'])"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"WR6CWADCcqyZ","executionInfo":{"status":"ok","timestamp":1610886247682,"user_tz":-60,"elapsed":7159,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"5410d243-1c6f-47f8-8bd5-46cd11ec190b"},"source":["simple_model.fit(x_train, y_train, validation_split=0.2, epochs=5, shuffle=True, batch_size=256)"],"execution_count":null,"outputs":[{"output_type":"stream","text":["Epoch 1/5\n","624/624 [==============================] - 2s 2ms/step - loss: 0.0393 - acc: 0.9834 - val_loss: 0.0034 - val_acc: 0.9994\n","Epoch 2/5\n","624/624 [==============================] - 1s 2ms/step - loss: 0.0028 - acc: 0.9995 - val_loss: 0.0032 - val_acc: 0.9995\n","Epoch 3/5\n","624/624 [==============================] - 1s 2ms/step - loss: 0.0027 - acc: 0.9994 - val_loss: 0.0028 - val_acc: 0.9994\n","Epoch 4/5\n","624/624 [==============================] - 1s 2ms/step - loss: 0.0023 - acc: 0.9994 - val_loss: 0.0029 - val_acc: 0.9995\n","Epoch 5/5\n","624/624 [==============================] - 1s 2ms/step - loss: 0.0021 - acc: 0.9995 - val_loss: 0.0034 - val_acc: 0.9992\n"],"name":"stdout"},{"output_type":"execute_result","data":{"text/plain":["<tensorflow.python.keras.callbacks.History at 0x7f7d92e1b4a8>"]},"metadata":{"tags":[]},"execution_count":14}]},{"cell_type":"code","metadata":{"id":"rR5DCu7OivPL"},"source":["def conf_matrix(predictions, class_types, y_test_argmax=False): \n","    ''' Plots conf. matrix and classification report '''\n","    if y_test_argmax==True:\n","      cm=confusion_matrix(np.argmax(y_test, axis=1), np.argmax(np.round(predictions), axis=1))\n","      cr=classification_report(np.argmax(y_test, axis=1),\n","                                np.argmax(np.round(predictions), axis=1), \n","                                target_names=[class_types[i] for i in range(len(class_types))])\n","      print(cr)\n","    else:\n","      cm=confusion_matrix(y_test, np.argmax(np.round(predictions), axis=1))\n","      print(\"Classification Report:\\n\")\n","      cr=classification_report(y_test,\n","                                  np.argmax(np.round(predictions), axis=1), \n","                                  target_names=[class_types[i] for i in range(len(class_types))])\n","      print(cr)\n","    plt.figure(figsize=(6,6))\n","    \n","    sns_hmp = sns.heatmap(cm, annot=True, xticklabels = [class_types[i] for i in range(len(class_types))], \n","                yticklabels = [class_types[i] for i in range(len(class_types))], fmt=\"d\", cmap=\"YlGnBu\")\n","    \n","    plt.xlabel('Pred Class')\n","    plt.ylabel('True Class')\n","    fig = sns_hmp.get_figure()\n","    # fig.savefig('/content/gdrive/My Drive/Colab Notebooks/resnet/heatmap.png', dpi=250)"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"CYiBnEI6d7En"},"source":["y_pred = simple_model.predict(x_test, batch_size=64)\n","\n","# y_pred_argmax = np.argmax(y_pred, axis=1)"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/","height":544},"id":"-raMLIzjjKQF","executionInfo":{"status":"ok","timestamp":1610886261837,"user_tz":-60,"elapsed":1012,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"4998a67e-6ebd-40c9-f903-454722674ec4"},"source":["conf_matrix(y_pred, ['Real', 'Fraud'], y_test_argmax=True)"],"execution_count":null,"outputs":[{"output_type":"stream","text":["              precision    recall  f1-score   support\n","\n","        Real       1.00      1.00      1.00     85295\n","       Fraud       0.87      0.47      0.61       148\n","\n","    accuracy                           1.00     85443\n","   macro avg       0.94      0.73      0.80     85443\n","weighted avg       1.00      1.00      1.00     85443\n","\n"],"name":"stdout"},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x432 with 2 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"vKF2WmD0ZU_c","executionInfo":{"status":"ok","timestamp":1610886291101,"user_tz":-60,"elapsed":670,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"f6116c40-ff63-4d26-8812-4eae1057540c"},"source":["print ('check first 3 y pred: ', y_pred[0:3])\n","print ('check first 3 y_test: ', y_test[0:3])\n","print ('check argmax of pred:',  np.argmax(np.round(y_pred[0:3]), axis=1))"],"execution_count":null,"outputs":[{"output_type":"stream","text":["check first 3 y pred:  [[9.9999940e-01 5.8592815e-07]\n"," [9.9996805e-01 3.1988002e-05]\n"," [9.9940896e-01 5.9102214e-04]]\n","check first 3 y_test:  [[1. 0.]\n"," [1. 0.]\n"," [1. 0.]]\n","check argmax of pred: [0 0 0]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"l-_llR0onUMA"},"source":["### Focal Loss: \n","\n","We will see that cross-entropy loss is a special case of Focal loss. \n","\n","Definition : \n","\n","$\\text{FL}(p_t) = - \\alpha _t \\, \\left(1-p_t\\right)^{\\gamma}\\, \\text{log}(p_t)$\n","\n","$$\\begin{equation}p_t = \\begin{cases}p\\, ; & \\text{if}\\, \\, y = 1 \\\\ 1-p\\, ; & \\text{otherwise} \\end{cases} \\end{equation}$$\n","\n","When $\\gamma = 0, \\alpha _t = 1$, then FL is Cross Entropy loss.  "]},{"cell_type":"code","metadata":{"id":"Aha7u0jxilUQ","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1612797021551,"user_tz":-60,"elapsed":692,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"30394a3a-39bc-4a12-8ae0-31e1794aba07"},"source":["## how does it affect ? \n","\n","y_example_true = [1.0]\n","y_example_pred = [0.90]\n","y_example_pred0 = [0.95]\n","y_example_pred1 = [0.20]\n","\n","binary_ce = tf.keras.losses.BinaryCrossentropy()\n","loss_close = binary_ce(y_example_true, y_example_pred).numpy()\n","loss_veryclose = binary_ce(y_example_true, y_example_pred0).numpy()\n","loss_far = binary_ce(y_example_true, y_example_pred1).numpy()\n","\n","print ('CE loss when pred is close to true: ', loss_close)\n","print ('CE loss when pred is very close to true: ', loss_veryclose)\n","print ('CE loss when pred is far from true: ', loss_far)\n","\n","focal_factor_close = (1.0-0.90)**2 ## (take gamma = 2, as in paper)\n","focal_factor_veryclose = (1.0-0.95)**2\n","focal_factor_far = (1.0-0.20)**2 ## ()\n","\n","print ('\\n')\n","\n","print ('focal loss when pred is close to true: ', loss_close*focal_factor_close)\n","print ('focal loss when pred is very close to true: ', loss_veryclose*focal_factor_veryclose)\n","\n","print ('focal loss when pred is far from true: ', loss_far*focal_factor_far)"],"execution_count":3,"outputs":[{"output_type":"stream","text":["CE loss when pred is close to true:  0.10536041\n","CE loss when pred is very close to true:  0.051293183\n","CE loss when pred is far from true:  1.6094373\n","\n","\n","focal loss when pred is close to true:  0.0010536041110754007\n","focal loss when pred is very close to true:  0.00012823295779526255\n","focal loss when pred is far from true:  1.0300399017333985\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"zcRSt5r-kr2a"},"source":["Penalize way more when prediction is far from true, i.e. focus on the wrongly classified samples"]},{"cell_type":"code","metadata":{"id":"xgu7BrRBj0wC"},"source":["import tensorflow_addons as tfa "],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"gRwEIAkbnT66","executionInfo":{"status":"ok","timestamp":1609247511538,"user_tz":-60,"elapsed":1100,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"c7f0b4e8-3552-4419-e390-4e86476183c7"},"source":["fl_tensorflow = tfa.losses.SigmoidFocalCrossEntropy(alpha=1.0, gamma=2.0)\n","fl_close = fl_tensorflow(y_example_true, y_example_pred).numpy()\n","fl_far = fl_tensorflow(y_example_true, y_example_pred1).numpy()\n","\n","print ('check values: ', fl_close, fl_far)"],"execution_count":null,"outputs":[{"output_type":"stream","text":["check values:  0.008925734 1.03004\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"wPhO3_8IoIRq"},"source":["Custom loss function "]},{"cell_type":"code","metadata":{"id":"0zT2OLFboE9N"},"source":["def focal_loss_custom(alpha, gamma):\n","  def binary_focal_loss(y_true, y_pred):\n","\n","    fl = tfa.losses.SigmoidFocalCrossEntropy(alpha=alpha, gamma=gamma)\n","    # pred_argmax = np.argmax(np.round(y_pred), axis=1)\n","    # pred_argmax_tensor = tf.convert_to_tensor(pred_argmax, dtype=tf.float32)\n","    # y_true_tensor = tf.convert_to_tensor(y_true, dtype=tf.float32)\n","    y_true_K = K.ones_like(y_true)\n","    focal_loss = fl(y_true, y_pred)\n","    # focal_loss = tf.reduce_mean(focal_loss)\n","    return focal_loss\n","  return binary_focal_loss  "],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"TLt0bUll1eAC"},"source":["simple_model1 = simple_model1()\n","simple_model1.compile(optimizer=Adam(learning_rate=5e-3), loss=focal_loss_custom(alpha=0.2, gamma=2.0), metrics=['acc'])"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"KQ7G9bHE4QpK","executionInfo":{"status":"ok","timestamp":1610886457806,"user_tz":-60,"elapsed":8044,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"d3cf7ef0-1acc-416d-9816-c10c037c154a"},"source":["simple_model1.fit(x_train, y_train, validation_split=0.2, epochs=5, shuffle=True, batch_size=256)"],"execution_count":null,"outputs":[{"output_type":"stream","text":["Epoch 1/5\n","624/624 [==============================] - 2s 2ms/step - loss: 0.0135 - acc: 0.9844 - val_loss: 0.0012 - val_acc: 0.9993\n","Epoch 2/5\n","624/624 [==============================] - 1s 2ms/step - loss: 8.6970e-04 - acc: 0.9994 - val_loss: 0.0013 - val_acc: 0.9991\n","Epoch 3/5\n","624/624 [==============================] - 1s 2ms/step - loss: 8.5765e-04 - acc: 0.9994 - val_loss: 0.0011 - val_acc: 0.9993\n","Epoch 4/5\n","624/624 [==============================] - 1s 2ms/step - loss: 7.4777e-04 - acc: 0.9994 - val_loss: 0.0012 - val_acc: 0.9993\n","Epoch 5/5\n","624/624 [==============================] - 1s 2ms/step - loss: 8.3562e-04 - acc: 0.9994 - val_loss: 0.0012 - val_acc: 0.9993\n"],"name":"stdout"},{"output_type":"execute_result","data":{"text/plain":["<tensorflow.python.keras.callbacks.History at 0x7f7a3594e390>"]},"metadata":{"tags":[]},"execution_count":15}]},{"cell_type":"code","metadata":{"id":"B5Bv6_si4W18"},"source":["y_pred = simple_model1.predict(x_test, batch_size=32)\n"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"o9k6Aaz2__VW","executionInfo":{"status":"ok","timestamp":1610886473060,"user_tz":-60,"elapsed":671,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"121462d0-b93a-4dbd-9e1c-642307543e6e"},"source":["confusion_matrix(np.argmax(y_test, axis=1), np.argmax(np.round(y_pred), axis=1))"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array([[85289,     6],\n","       [   47,   101]])"]},"metadata":{"tags":[]},"execution_count":17}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/","height":544},"id":"oB2_zHeix_6C","executionInfo":{"status":"ok","timestamp":1610886477687,"user_tz":-60,"elapsed":1092,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"184d2014-df27-4daa-9572-30b564a0b328"},"source":["conf_matrix(y_pred, ['Real', 'Fraud'], y_test_argmax=True)"],"execution_count":null,"outputs":[{"output_type":"stream","text":["              precision    recall  f1-score   support\n","\n","        Real       1.00      1.00      1.00     85295\n","       Fraud       0.94      0.68      0.79       148\n","\n","    accuracy                           1.00     85443\n","   macro avg       0.97      0.84      0.90     85443\n","weighted avg       1.00      1.00      1.00     85443\n","\n"],"name":"stdout"},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x432 with 2 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"code","metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"mplD53m9EWG5","executionInfo":{"status":"ok","timestamp":1610886577512,"user_tz":-60,"elapsed":661,"user":{"displayName":"Swap vi","photoUrl":"","userId":"01936573407644251994"}},"outputId":"4c6b85ca-92bf-4193-d853-f93425467cad"},"source":["print ('check first 3 ypred: ', y_pred[0:4])\n","print ('check first 3 ytest: ', y_test[0:4])\n","\n","print (np.argmax(np.round(y_pred[0:4]), axis=1))\n","\n","print (tf.cast(y_test[0:4], dtype=tf.float32))\n","print (K.ones_like(y_test[0:4]))"],"execution_count":null,"outputs":[{"output_type":"stream","text":["check first 3 ypred:  [[0.9667234  0.03327663]\n"," [0.9822015  0.01779843]\n"," [0.99318063 0.00681934]\n"," [0.9779823  0.02201766]]\n","check first 3 ytest:  [[1. 0.]\n"," [1. 0.]\n"," [1. 0.]\n"," [1. 0.]]\n","[0 0 0 0]\n","tf.Tensor(\n","[[1. 0.]\n"," [1. 0.]\n"," [1. 0.]\n"," [1. 0.]], shape=(4, 2), dtype=float32)\n","tf.Tensor(\n","[[1. 1.]\n"," [1. 1.]\n"," [1. 1.]\n"," [1. 1.]], shape=(4, 2), dtype=float32)\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"iK73PVatOyWk"},"source":[""],"execution_count":null,"outputs":[]}]}