Implementation Neural Network

nn_basic_implementation.json

{"cells": [{"cell_type": "code", "metadata": {"trusted": true, "collapsed": false}, "execution_count": null, "outputs": [], "source": "import numpy as np\n\ndef loadDataSet():\n    X = np.array([[1,1,1],[1,0,1],[1,1,0],[1,0,0]])\n    y = np.array([[0],[1],[1],[0]])\n    return X,y\n\ndef costError(w_1,w_2,currentIter):\n    z_1 = X.dot(w_1.T)\n    a_1 = np.hstack((np.ones((n_observations,1)),sigmoid(z_1)))\n    \n    z_2 = a_1.dot(w_2.T)\n    a_2 = sigmoid(z_2)\n    \n    sumError = np.mean((y - a_2)**2)\n    print(\"Error : %f , Iter: %d\"%(sumError,currentIter))\n    \n    \ndef sigmoid(z,derivate=False):\n    if derivate:\n        return sigmoid(z) * (1 - sigmoid(z))\n    return 1 / (1 + np.e**-z)\n\ndef classify(w_1,w_2,input_test):\n    z_1 = input_test.dot(w_1.T)\n    a_1 = np.hstack((np.ones((1,1)),sigmoid(z_1)))\n    \n    z_2 = a_1.dot(w_2.T)\n    a_2 = sigmoid(z_2)\n    \n    if a_2 >= 0.5:\n        print(\"Classification : 1 , Prob: %f\"%a_2)\n    else:\n        print(\"Classification : 0 , Prob: %f\"%a_2)\n\nX,y = loadDataSet()\nn_observations,n_features = X.shape\nweights_1 = np.random.rand(2,3)\nweights_2 = np.random.rand(1,3)\n\nmaxIter = 5000\nfor currentIter in range(maxIter):\n    costError(weights_1,weights_2,currentIter)\n    \n    #FeedFodward\n    z_1 = X.dot(weights_1.T)\n    a_1 = np.hstack((np.ones((n_observations,1)),sigmoid(z_1)))\n    \n    z_2 = a_1.dot(weights_2.T)\n    a_2 = sigmoid(z_2)\n    \n    #BackPropagation\n    d_2 = (y - a_2) #* sigmoid(z_2,derivate=True)\n    d_1 = ((d_2.dot(weights_2))[:,1:]) * sigmoid(z_1,derivate=True)\n    \n    alpha = 1 #Learning Rate\n    weights_2 = weights_2 + alpha * d_2.T.dot(a_1)\n    weights_1 = weights_1 + alpha * d_1.T.dot(X)\n    \nclassify(weights_1,weights_2,np.array([[1,1,1]])) \nclassify(weights_1,weights_2,np.array([[1,0,1]])) \nclassify(weights_1,weights_2,np.array([[1,1,0]])) \nclassify(weights_1,weights_2,np.array([[1,0,0]])) \n    "}], "metadata": {"kernelspec": {"display_name": "Python 3", "name": "python3", "language": "python"}, "language_info": {"name": "python", "file_extension": ".py", "version": "3.4.3", "pygments_lexer": "ipython3", "codemirror_mode": {"name": "ipython", "version": 3}, "mimetype": "text/x-python", "nbconvert_exporter": "python"}}, "nbformat_minor": 0, "nbformat": 4}