glove_vecs.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "vector1 = np.random.rand(200)\n",
    "vector2 = np.random.rand(200)\n",
    "vector = np.random.rand(300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Usefull things for ML/DL"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#   https://nlp.stanford.edu/projects/glove/        - download pretrained glove vector word representation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#   https://nlp.stanford.edu/pubs/glove.pdf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# L2 norm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "L2norm = np.linalg.norm(vector)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9.846690841270856\n"
     ]
    }
   ],
   "source": [
    "print(L2norm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Matrix multiplication"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "dot = np.dot(vector1, vector2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "51.78968201654798\n"
     ]
    }
   ],
   "source": [
    "print(dot)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Absolute value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "abs_vector1 = np.absolute(vector1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Cosine_similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "cos_similarity = np.dot(vector1, vector2) / np.dot( np.linalg.norm(vector1), np.linalg.norm(vector2) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7673235450051022\n"
     ]
    }
   ],
   "source": [
    "print(cos_similarity)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cosine_similarity(vactor1,vector2):\n",
    "    # u, v - vectors of shape (n,)\n",
    "    return np.dot(vector1, vector2) / np.dot( np.linalg.norm(vector1), np.linalg.norm(vector2) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Activation functions:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Sigmoid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sigmoid(x):\n",
    "    return 1. / (1 + np.exp(-x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# dSigmoid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def dsigmoid(x): \n",
    "    # x == sigmoid(x)\n",
    "    return x * (1. - x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ReLU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def ReLU(x):\n",
    "    return np.maximum(0,x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# dReLU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def dReLU(x):\n",
    "    return 1. * (x > 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# LeakyReLU - allows a small, non zero gradient when the unit is not active"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "def LeakyReLU(x, epsilon=0.01):\n",
    "    return np.maximum(epsilon * x, x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# dLeakyReLU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def dLeakyReLU(x, alpha=0.01):\n",
    "    dx = np.ones_like(x)\n",
    "    dx[x < 0] = alpha\n",
    "    return dx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TanH"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "def tanh(x):\n",
    "    return np.tanh(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def dtanh(x):\n",
    "    return 1. - x * x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "#----"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = np.linspace(-3, 3, 1000)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_sigmoid = sigmoid(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_ReLU = ReLU(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_LeakyReLU = LeakyReLU(x,0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_tanh = tanh(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(x,y_sigmoid,label=\"sigmoid\",color=\"red\",linewidth=2)  \n",
    "\n",
    "plt.plot(x,y_ReLU,label=\"ReLU\",color=\"blue\",linewidth=2)  \n",
    "\n",
    "plt.plot(x,y_LeakyReLU,label=\"LeakyReLU\",color=\"pink\",linewidth=2)  \n",
    "\n",
    "plt.plot(x,y_tanh,label=\"tanh\",color=\"orange\",linewidth=2)  \n",
    "\n",
    "plt.xlabel(\"x\")  \n",
    "plt.ylabel(\"y\")  \n",
    "plt.title(\"Activation functions\") \n",
    "#plt.xscale('log')\n",
    "#plt.yscale('log')\n",
    "\n",
    "legend = plt.legend(loc='upper left', shadow=True)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Softmax"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "def softmax(x):\n",
    "    ex = np.exp(x-np.max(x))\n",
    "    return ex / ex.sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "#----"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# How to read glove vector word representation file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_glove_vecs(glove_file):\n",
    "   with open(glove_file, 'r') as f:\n",
    "       words = set()\n",
    "       word_to_vec_map = {}\n",
    "       for line in f:\n",
    "           line = line.strip().split()\n",
    "           curr_word = line[0]\n",
    "           words.add(curr_word)\n",
    "           word_to_vec_map[curr_word] = np.array(line[1:], dtype=np.float64)\n",
    "   return words, word_to_vec_map"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7673235450051022\n"
     ]
    }
   ],
   "source": [
    "print(cosine_similarity(vector1,vector2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "glove_file = '/Users/tpalczew/Downloads/glove.6B.50d.txt'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "words, word_to_vec_map = read_glove_vecs(glove_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.50451   0.68607  -0.59517  -0.022801  0.60046  -0.13498  -0.08813\n",
      "  0.47377  -0.61798  -0.31012  -0.076666  1.493    -0.034189 -0.98173\n",
      "  0.68229   0.81722  -0.51874  -0.31503  -0.55809   0.66421   0.1961\n",
      " -0.13495  -0.11476  -0.30344   0.41177  -2.223    -1.0756   -1.0783\n",
      " -0.34354   0.33505   1.9927   -0.04234  -0.64319   0.71125   0.49159\n",
      "  0.16754   0.34344  -0.25663  -0.8523    0.1661    0.40102   1.1685\n",
      " -1.0137   -0.21585  -0.15155   0.78321  -0.91241  -1.6106   -0.64426\n",
      " -0.51042 ]\n"
     ]
    }
   ],
   "source": [
    "print(word_to_vec_map[\"king\"])"
   ]
  },
  {
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}