ELM example.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from tensorflow.keras.datasets import mnist\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from matplotlib import pyplot as plt\n",
    "import random\n",
    "from model import ELM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Network Settings\n",
    "num_classes = 10\n",
    "num_hidden_layers = 512\n",
    "input_length = 28**2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load MNIST Dataset\n",
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
    "\n",
    "# Process images into input vectors\n",
    "# each mnist image is a 28x28 picture with value ranges between 0 and 255\n",
    "x_train = x_train.astype(np.float32) / 255.\n",
    "x_train = x_train.reshape(-1, input_length)\n",
    "x_test = x_test.astype(np.float32) / 255.\n",
    "x_test = x_test.reshape(-1, input_length)\n",
    "\n",
    "# converts [1,2] into [[0,1,0], [0,0,1]]\n",
    "y_train = to_categorical(y_train, num_classes).astype(np.float32)\n",
    "y_test = to_categorical(y_test, num_classes).astype(np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Bias shape: (512,)\n",
      "W shape: (784, 512)\n",
      "Beta shape: (512, 10)\n"
     ]
    }
   ],
   "source": [
    "# create instance of our model\n",
    "model = ELM(\n",
    "    input_length,\n",
    "    num_hidden_layers,\n",
    "    num_classes\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train time: 3.594151735305786\n",
      "train loss: 0.014729\n",
      "train acc: 0.910000\n"
     ]
    }
   ],
   "source": [
    "# Train\n",
    "model.fit(x_train, y_train, display_time=True)\n",
    "train_loss, train_acc = model.evaluate(x_train, y_train)\n",
    "print('train loss: %f' % train_loss)\n",
    "print('train acc: %f' % train_acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val loss: 0.014554\n",
      "val acc: 0.911100\n"
     ]
    }
   ],
   "source": [
    "# Validation\n",
    "val_loss, val_acc = model.evaluate(x_test, y_test)\n",
    "print('val loss: %f' % val_loss)\n",
    "print('val acc: %f' % val_acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1332x756 with 5 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Predictions\n",
    "test_data = x_test[np.random.choice(len(x_test), size=5, replace=False)]\n",
    "predictions = model(test_data)\n",
    "\n",
    "fig, axs = plt.subplots(1, 5)\n",
    "fig.set_size_inches(18.5, 10.5)\n",
    "for idx, image in enumerate(test_data):\n",
    "    image = np.array(image, dtype='float')\n",
    "    pixels = image.reshape((28, 28))\n",
    "    axs[idx].imshow(pixels, cmap='gray')\n",
    "    axs[idx].axes.get_yaxis().set_visible(False)\n",
    "    axs[idx].title.set_text(f'Predicted {np.argmax(predictions[idx])}')\n",
    "    axs[idx].set_xlabel(f'With prob: {str(round(np.max(predictions[idx]), 2))}')\n",
    "    \n",
    "    \n",
    "plt.show()"
   ]
  },
  {
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}