NN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#sigmoid activation function which is non linear\n",
    "def sigmoid(x):\n",
    "    return 1/(1+np.exp(-1*x))\n",
    "def sigmoid_derivative(x):\n",
    "    return x*(1-x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "#architecture: 1 hidden layer with 4 neurons, outpur layer with 1 neuron\n",
    "class NeuralNetwork:\n",
    "    def __init__(self, x,y):\n",
    "        self.input = x\n",
    "        #random initialization of weights for neurons\n",
    "        self.weights1= np.random.rand(self.input.shape[1],4)\n",
    "        self.weights2 = np.random.rand(4,1)\n",
    "        self.y = y\n",
    "        self.output = np. zeros(y.shape)\n",
    "        \n",
    "    def feedforward(self):\n",
    "        self.layer1 = sigmoid(np.dot(self.input, self.weights1))\n",
    "        self.layer2 = sigmoid(np.dot(self.layer1, self.weights2))\n",
    "        return self.layer2\n",
    "        \n",
    "    def backprop(self):\n",
    "        #using chain rule updating weights using gradient descent optimizer\n",
    "        out_der = 2*(self.y -self.output)*sigmoid_derivative(self.output)\n",
    "        d_weights2 = np.dot(self.layer1.T,out_der)\n",
    "        hid_der = np.dot(out_der, self.weights2.T)*sigmoid_derivative(self.layer1)\n",
    "        d_weights1 = np.dot(self.input.T, hid_der)\n",
    "    \n",
    "        self.weights1 += d_weights1\n",
    "        self.weights2 += d_weights2\n",
    "        \n",
    "        print(d_weights1.shape)\n",
    "        print(d_weights2.shape)\n",
    "\n",
    "    def train(self):\n",
    "        self.output = self.feedforward()\n",
    "        self.backprop()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "X=np.array([[0,0],[0,1],[1,0],[1,1]])\n",
    "y=np.array([0,1,1,0]).reshape(4,1)\n",
    "nn=NeuralNetwork(X,y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2, 4)\n",
      "(4, 1)\n",
      "[[-0.7443047 ]\n",
      " [ 0.1852634 ]\n",
      " [ 0.20800095]\n",
      " [-0.84269221]]\n"
     ]
    }
   ],
   "source": [
    "for i in range(1):\n",
    "    nn.train()\n",
    "    if(i%500==0):\n",
    "        print(nn.y-nn.output)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(4,)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a=np.array([1,2,3,4]).reshape(4,1)\n",
    "a*(1-a)\n",
    "y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.05365522],\n",
       "       [0.95578176],\n",
       "       [0.9548636 ],\n",
       "       [0.0388071 ]])"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nn.output"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.15rc1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}