A barebone introduction to Keras

learn-keras-1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from random import randint\n",
    "from sklearn.utils import shuffle\n",
    "from sklearn.preprocessing import MinMaxScaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_labels = []\n",
    "train_samples = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# initializing ages and side-effect for 5% of people\n",
    "for i in range(50):\n",
    "    random_younger = randint(13,64)\n",
    "    train_samples.append(random_younger)\n",
    "    train_labels.append(1) # side-effect\n",
    "    \n",
    "    random_older = randint(65,100)\n",
    "    train_samples.append(random_older)\n",
    "    train_labels.append(0) # no side-effect"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# initializing ages and appearance of side-effects for 95% people\n",
    "for i in range(1000):\n",
    "    random_younger = randint(13, 64)\n",
    "    train_samples.append(random_younger)\n",
    "    train_labels.append(0)\n",
    "    \n",
    "    random_older = randint(65, 100)\n",
    "    train_samples.append(random_older)\n",
    "    train_labels.append(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[44, 85, 53, 82, 45]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_samples[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[1, 0, 1, 0, 1]"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_labels[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_labels = np.array(train_labels)\n",
    "train_samples = np.array(train_samples)\n",
    "train_labels, train_samples = shuffle(train_labels, train_samples)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "scaler = MinMaxScaler(feature_range=(0,1))\n",
    "scaled_train_samples = scaler.fit_transform(train_samples.reshape(-1, 1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.27586207],\n",
       "       [0.22988506],\n",
       "       [0.11494253],\n",
       "       [0.03448276],\n",
       "       [0.56321839]])"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scaled_train_samples[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Activation, Dense\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.metrics import categorical_crossentropy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential([\n",
    "    Dense(units=16, input_shape=(1,), activation='relu'),\n",
    "    Dense(units=32, activation='relu'),\n",
    "    Dense(units=2, activation='softmax')\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 16)                32        \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                544       \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 2)                 66        \n",
      "=================================================================\n",
      "Total params: 642\n",
      "Trainable params: 642\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer=Adam(learning_rate=0.0001), loss='sparse_categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "189/189 - 1s - loss: 0.7034 - accuracy: 0.5079 - val_loss: 0.6826 - val_accuracy: 0.5429\n",
      "Epoch 2/30\n",
      "189/189 - 0s - loss: 0.6606 - accuracy: 0.6307 - val_loss: 0.6481 - val_accuracy: 0.6667\n",
      "Epoch 3/30\n",
      "189/189 - 0s - loss: 0.6235 - accuracy: 0.7148 - val_loss: 0.6166 - val_accuracy: 0.7000\n",
      "Epoch 4/30\n",
      "189/189 - 0s - loss: 0.5863 - accuracy: 0.7619 - val_loss: 0.5857 - val_accuracy: 0.7476\n",
      "Epoch 5/30\n",
      "189/189 - 0s - loss: 0.5496 - accuracy: 0.8143 - val_loss: 0.5546 - val_accuracy: 0.7571\n",
      "Epoch 6/30\n",
      "189/189 - 1s - loss: 0.5134 - accuracy: 0.8317 - val_loss: 0.5239 - val_accuracy: 0.8000\n",
      "Epoch 7/30\n",
      "189/189 - 1s - loss: 0.4783 - accuracy: 0.8598 - val_loss: 0.4952 - val_accuracy: 0.8333\n",
      "Epoch 8/30\n",
      "189/189 - 1s - loss: 0.4455 - accuracy: 0.8709 - val_loss: 0.4683 - val_accuracy: 0.8524\n",
      "Epoch 9/30\n",
      "189/189 - 1s - loss: 0.4155 - accuracy: 0.8862 - val_loss: 0.4439 - val_accuracy: 0.8714\n",
      "Epoch 10/30\n",
      "189/189 - 0s - loss: 0.3882 - accuracy: 0.8968 - val_loss: 0.4218 - val_accuracy: 0.8857\n",
      "Epoch 11/30\n",
      "189/189 - 0s - loss: 0.3646 - accuracy: 0.9106 - val_loss: 0.4039 - val_accuracy: 0.8905\n",
      "Epoch 12/30\n",
      "189/189 - 1s - loss: 0.3456 - accuracy: 0.9111 - val_loss: 0.3892 - val_accuracy: 0.9048\n",
      "Epoch 13/30\n",
      "189/189 - 1s - loss: 0.3302 - accuracy: 0.9206 - val_loss: 0.3773 - val_accuracy: 0.9048\n",
      "Epoch 14/30\n",
      "189/189 - 0s - loss: 0.3172 - accuracy: 0.9201 - val_loss: 0.3673 - val_accuracy: 0.9000\n",
      "Epoch 15/30\n",
      "189/189 - 0s - loss: 0.3070 - accuracy: 0.9280 - val_loss: 0.3591 - val_accuracy: 0.9000\n",
      "Epoch 16/30\n",
      "189/189 - 0s - loss: 0.2984 - accuracy: 0.9275 - val_loss: 0.3525 - val_accuracy: 0.9000\n",
      "Epoch 17/30\n",
      "189/189 - 0s - loss: 0.2917 - accuracy: 0.9312 - val_loss: 0.3472 - val_accuracy: 0.9000\n",
      "Epoch 18/30\n",
      "189/189 - 1s - loss: 0.2861 - accuracy: 0.9312 - val_loss: 0.3430 - val_accuracy: 0.9000\n",
      "Epoch 19/30\n",
      "189/189 - 1s - loss: 0.2815 - accuracy: 0.9344 - val_loss: 0.3396 - val_accuracy: 0.9000\n",
      "Epoch 20/30\n",
      "189/189 - 1s - loss: 0.2775 - accuracy: 0.9333 - val_loss: 0.3356 - val_accuracy: 0.9095\n",
      "Epoch 21/30\n",
      "189/189 - 1s - loss: 0.2743 - accuracy: 0.9344 - val_loss: 0.3331 - val_accuracy: 0.9095\n",
      "Epoch 22/30\n",
      "189/189 - 1s - loss: 0.2714 - accuracy: 0.9328 - val_loss: 0.3309 - val_accuracy: 0.9095\n",
      "Epoch 23/30\n",
      "189/189 - 0s - loss: 0.2692 - accuracy: 0.9365 - val_loss: 0.3291 - val_accuracy: 0.9095\n",
      "Epoch 24/30\n",
      "189/189 - 0s - loss: 0.2671 - accuracy: 0.9354 - val_loss: 0.3274 - val_accuracy: 0.9095\n",
      "Epoch 25/30\n",
      "189/189 - 1s - loss: 0.2653 - accuracy: 0.9360 - val_loss: 0.3253 - val_accuracy: 0.9095\n",
      "Epoch 26/30\n",
      "189/189 - 1s - loss: 0.2635 - accuracy: 0.9365 - val_loss: 0.3239 - val_accuracy: 0.9095\n",
      "Epoch 27/30\n",
      "189/189 - 0s - loss: 0.2620 - accuracy: 0.9365 - val_loss: 0.3227 - val_accuracy: 0.9095\n",
      "Epoch 28/30\n",
      "189/189 - 0s - loss: 0.2607 - accuracy: 0.9365 - val_loss: 0.3216 - val_accuracy: 0.9095\n",
      "Epoch 29/30\n",
      "189/189 - 0s - loss: 0.2595 - accuracy: 0.9365 - val_loss: 0.3201 - val_accuracy: 0.9095\n",
      "Epoch 30/30\n",
      "189/189 - 0s - loss: 0.2583 - accuracy: 0.9386 - val_loss: 0.3190 - val_accuracy: 0.9095\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x7f2eefb01880>"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(x=scaled_train_samples, y=train_labels, validation_split=0.1, batch_size=10, epochs=30, shuffle=True, verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}