Created
April 29, 2021 04:39
-
-
Save alexcpn/4d13ab72c8dcd97d81c296474621197a to your computer and use it in GitHub Desktop.
explain_neuralnet_asifiama_ schoolkid_2.ipynb
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| { | |
| "nbformat": 4, | |
| "nbformat_minor": 0, | |
| "metadata": { | |
| "colab": { | |
| "name": "explain_neuralnet_asifiama_ schoolkid_2.ipynb", | |
| "provenance": [], | |
| "collapsed_sections": [], | |
| "include_colab_link": true | |
| }, | |
| "kernelspec": { | |
| "name": "python3", | |
| "display_name": "Python 3" | |
| } | |
| }, | |
| "cells": [ | |
| { | |
| "cell_type": "markdown", | |
| "metadata": { | |
| "id": "view-in-github", | |
| "colab_type": "text" | |
| }, | |
| "source": [ | |
| "<a href=\"https://colab.research.google.com/gist/alexcpn/4d13ab72c8dcd97d81c296474621197a/explain_neuralnet_asifiama_-schoolkid_2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>" | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": { | |
| "id": "WUMcoquq5pRe" | |
| }, | |
| "source": [ | |
| "The below is the neural net we want to model; Taking inputs x0 to x4 and giving an output " | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": { | |
| "id": "sC9m5k2349vT" | |
| }, | |
| "source": [ | |
| "" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "metadata": { | |
| "id": "dihdwBm62yKb" | |
| }, | |
| "source": [ | |
| "# This x defines our Input; or in Neural Network terms, the training set\n", | |
| "\n", | |
| "x = np.array(\n", | |
| " [\n", | |
| " [0,0,1],\n", | |
| " [0,1,1],\n", | |
| " [1,0,1],\n", | |
| " [1,1,1]\n", | |
| " ])\n" | |
| ], | |
| "execution_count": null, | |
| "outputs": [] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "metadata": { | |
| "id": "isn3yMg36_rd" | |
| }, | |
| "source": [ | |
| "# and let y be the desired ouput \n", | |
| "# Basically the first row in y represents the exected ouput for the first row of the training set.\n", | |
| "# That is the output becomes1 only when two of the inputs\n", | |
| "# becomes one\n", | |
| "y = np.array(\n", | |
| " [\n", | |
| " [0],\n", | |
| " [1],\n", | |
| " [1],\n", | |
| " [0]\n", | |
| " ])" | |
| ], | |
| "execution_count": null, | |
| "outputs": [] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": { | |
| "id": "69cIvPwAg-Fz" | |
| }, | |
| "source": [ | |
| "Now we need to define the weights of the layers of our neural network.\n", | |
| "Note that the input is of shape $4*3$. So the first layer should be of shape that we can take dot product of.\n", | |
| "Let us use $3*4$. The shape of the dot product woud be $4*4$ and the next layer weight should be comptaible with it. Let's take the next layer weights as size $4*1$." | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "metadata": { | |
| "id": "lDXCEUlezpgd" | |
| }, | |
| "source": [ | |
| "np.random.seed(1)\n", | |
| "\n", | |
| "# randomly initialize our weights with mean 0\n", | |
| "\n", | |
| "weight1 = np.random.random((3,4)) \n", | |
| "weight2 = np.random.random((4,1)) \n" | |
| ], | |
| "execution_count": null, | |
| "outputs": [] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "metadata": { | |
| "id": "17fD9mvZCHjb", | |
| "colab": { | |
| "base_uri": "https://localhost:8080/" | |
| }, | |
| "outputId": "416b96d4-e0ef-4bf8-f41b-6e6638e4f072" | |
| }, | |
| "source": [ | |
| "#---------------------------------------------------------------\n", | |
| "# Boiler plate code for calculating Sigmoid, derivative etc\n", | |
| "#---------------------------------------------------------------\n", | |
| "\n", | |
| "import numpy as np\n", | |
| "# seed random numbers to make calculation deterministic \n", | |
| "np.random.seed(1)\n", | |
| "\n", | |
| "# pretty print numpy array\n", | |
| "np.set_printoptions(formatter={'float': '{: 0.3f}'.format})\n", | |
| "\n", | |
| "# let us code our sigmoid funciton\n", | |
| "def sigmoid(x):\n", | |
| " return 1/(1+np.exp(-x))\n", | |
| "\n", | |
| "# let us add a method that takes the derivative of x as well\n", | |
| "def derv_sigmoid(x):\n", | |
| " return sigmoid(x)*(1-sigmoid(x))\n", | |
| "\n", | |
| "#---------------------------------------------------------------\n", | |
| "\n", | |
| "# Two layered NW. Using from (1) and the equations we derived as explanaionns\n", | |
| "# (1) http://iamtrask.github.io/2015/07/12/basic-python-network/\n", | |
| "#---------------------------------------------------------------\n", | |
| "\n", | |
| "# set learning rate as 1 for this toy example\n", | |
| "learningRate = 1\n", | |
| "\n", | |
| "# input x, also used as the training set here\n", | |
| "x = np.array([ [0,0,1],[0,1,1],[1,0,1],[1,1,1] ])\n", | |
| "\n", | |
| "# desired output for each of the training set above\n", | |
| "y = np.array([[0,1,1,0]]).T\n", | |
| "\n", | |
| "# Explanaiton - as long as input has two ones, but not three, ouput is One\n", | |
| "\"\"\"\n", | |
| "Input [0,0,1] Output = 0\n", | |
| "Input [0,1,1] Output = 1\n", | |
| "Input [1,0,1] Output = 1\n", | |
| "Input [1,1,1] Output = 0\n", | |
| "\"\"\"\n", | |
| "\n", | |
| "# Randomly initalised weights\n", | |
| "weight1 = np.random.random((3,4)) \n", | |
| "weight2 = np.random.random((4,1)) \n", | |
| "\n", | |
| "# Activation to layer 0 is taken as input x\n", | |
| "a0 = x\n", | |
| "\n", | |
| "iterations = 1000\n", | |
| "for iter in range(0,iterations):\n", | |
| "\n", | |
| " # Forward pass - Straight Forward\n", | |
| " z1= np.dot(x,weight1)\n", | |
| " a1 = sigmoid(z1) \n", | |
| " z2= np.dot(a1,weight2)\n", | |
| " a2 = sigmoid(z2) \n", | |
| " if iter == 0:\n", | |
| " print(\"Intial Ouput \\n\",a2)\n", | |
| "\n", | |
| " # Backward Pass - Backpropagation \n", | |
| " delta2 = (a2-y)\n", | |
| " #---------------------------------------------------------------\n", | |
| " # Calcluating change of Cost/Loss wrto weight of 2nd/last layer\n", | |
| " # Eq (A) ---> dC_dw2 = delta2*derv_sigmoid(z2)*a1.T\n", | |
| " #---------------------------------------------------------------\n", | |
| "\n", | |
| " dC_dw2_1 = delta2*derv_sigmoid(z2) \n", | |
| " dC_dw2 = a1.T.dot(dC_dw2_1)\n", | |
| " \n", | |
| " #---------------------------------------------------------------\n", | |
| " # Calcluating change of Cost/Loss wrto weight of 2nd/last layer\n", | |
| " # Eq (B)---> dC_dw1 = derv_sigmoid(z1)*delta2*derv_sigmoid(z2)*weight2*a0.T\n", | |
| " # dC_dw1 = derv_sigmoid(z1)*dC_dw2*weight2_1*a0.T\n", | |
| " #---------------------------------------------------------------\n", | |
| "\n", | |
| " dC_dw1 = np.multiply(dC_dw2_1,weight2.T) * derv_sigmoid(z1)\n", | |
| " dC_dw1 = a0.T.dot(dC_dw1)\n", | |
| "\n", | |
| " #---------------------------------------------------------------\n", | |
| " #Gradinent descent\n", | |
| " #---------------------------------------------------------------\n", | |
| " \n", | |
| " weight2 = weight2 - learningRate*(dC_dw2)\n", | |
| " weight1 = weight1 - learningRate*(dC_dw1)\n", | |
| "\n", | |
| "\n", | |
| "print(\"New ouput\",a2)\n", | |
| "\n", | |
| "#---------------------------------------------------------------\n", | |
| "# Training is done, weight2 and weight2 are primed for output y\n", | |
| "#---------------------------------------------------------------\n", | |
| "\n", | |
| "# Lets test out, two ones in input and one zero, ouput should be One\n", | |
| "x = np.array([[1,0,1]])\n", | |
| "z1= np.dot(x,weight1)\n", | |
| "a1 = sigmoid(z1) \n", | |
| "z2= np.dot(a1,weight2)\n", | |
| "a2 = sigmoid(z2) \n", | |
| "print(\"Ouput after Training is \\n\",a2)\n", | |
| " \n" | |
| ], | |
| "execution_count": 3, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": [ | |
| "Intial Ouput \n", | |
| " [[ 0.758]\n", | |
| " [ 0.771]\n", | |
| " [ 0.791]\n", | |
| " [ 0.801]]\n", | |
| "New ouput [[ 0.028]\n", | |
| " [ 0.925]\n", | |
| " [ 0.925]\n", | |
| " [ 0.090]]\n", | |
| "Ouput after Training is \n", | |
| " [[ 0.925]]\n" | |
| ], | |
| "name": "stdout" | |
| } | |
| ] | |
| }, | |
| { | |
| "cell_type": "markdown", | |
| "metadata": { | |
| "id": "Rg-nU_ZNrD-s" | |
| }, | |
| "source": [ | |
| "Lets model the above via Tensorflow framework" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "metadata": { | |
| "colab": { | |
| "base_uri": "https://localhost:8080/", | |
| "height": 642 | |
| }, | |
| "id": "-1SsJY0mGWJm", | |
| "outputId": "85fd33a3-f225-45b8-a927-a85b0454faf0" | |
| }, | |
| "source": [ | |
| "import tensorflow as tf\n", | |
| "from tensorflow import keras\n", | |
| "from tensorflow.keras import layers\n", | |
| "import numpy as np\n", | |
| "from matplotlib import pyplot\n", | |
| "\n", | |
| "# Define Sequential model with 3 layers\n", | |
| "# https://keras.io/api/layers/core_layers/dense/\n", | |
| "\n", | |
| "model = keras.Sequential()\n", | |
| "model.add(layers.Dense(3, activation=\"sigmoid\", name=\"layer1\",input_shape=(4,3)))\n", | |
| "model.add(layers.Dense(4, activation=\"sigmoid\", name=\"layer2\"))\n", | |
| "model.add(layers.Dense(1, activation=\"sigmoid\", name=\"layer3\"))\n", | |
| "\n", | |
| "opt = tf.keras.optimizers.SGD(learning_rate=1.0)\n", | |
| "model.compile(optimizer=opt, loss='mse',metrics=[['accuracy'], ['mse']])\n", | |
| "# Call model on a test input\n", | |
| "x = np.array(\n", | |
| " [\n", | |
| " [0,0,1],\n", | |
| " [0,1,1],\n", | |
| " [1,0,1],\n", | |
| " [1,1,1]\n", | |
| " ])\n", | |
| "y = np.array([[0,1,1,0]]).T\n", | |
| "model.output_shape\n", | |
| "model.summary()\n", | |
| "history =model.fit(x, y, epochs=1000, batch_size=1,verbose=0)\n", | |
| "pyplot.plot(history.history['accuracy'])\n", | |
| "pyplot.show()\n", | |
| "pred = model.predict(x) \n", | |
| "print(\"prediction\",pred)" | |
| ], | |
| "execution_count": 2, | |
| "outputs": [ | |
| { | |
| "output_type": "stream", | |
| "text": [ | |
| "Model: \"sequential_1\"\n", | |
| "_________________________________________________________________\n", | |
| "Layer (type) Output Shape Param # \n", | |
| "=================================================================\n", | |
| "layer1 (Dense) (None, 4, 3) 12 \n", | |
| "_________________________________________________________________\n", | |
| "layer2 (Dense) (None, 4, 4) 16 \n", | |
| "_________________________________________________________________\n", | |
| "layer3 (Dense) (None, 4, 1) 5 \n", | |
| "=================================================================\n", | |
| "Total params: 33\n", | |
| "Trainable params: 33\n", | |
| "Non-trainable params: 0\n", | |
| "_________________________________________________________________\n", | |
| "WARNING:tensorflow:Model was constructed with shape (None, 4, 3) for input KerasTensor(type_spec=TensorSpec(shape=(None, 4, 3), dtype=tf.float32, name='layer1_input'), name='layer1_input', description=\"created by layer 'layer1_input'\"), but it was called on an input with incompatible shape (1, 3).\n", | |
| "WARNING:tensorflow:Model was constructed with shape (None, 4, 3) for input KerasTensor(type_spec=TensorSpec(shape=(None, 4, 3), dtype=tf.float32, name='layer1_input'), name='layer1_input', description=\"created by layer 'layer1_input'\"), but it was called on an input with incompatible shape (1, 3).\n" | |
| ], | |
| "name": "stdout" | |
| }, | |
| { | |
| "output_type": "display_data", | |
| "data": { | |
| "image/png": "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\n", | |
| "text/plain": [ | |
| "<Figure size 432x288 with 1 Axes>" | |
| ] | |
| }, | |
| "metadata": { | |
| "tags": [], | |
| "needs_background": "light" | |
| } | |
| }, | |
| { | |
| "output_type": "stream", | |
| "text": [ | |
| "WARNING:tensorflow:Model was constructed with shape (None, 4, 3) for input KerasTensor(type_spec=TensorSpec(shape=(None, 4, 3), dtype=tf.float32, name='layer1_input'), name='layer1_input', description=\"created by layer 'layer1_input'\"), but it was called on an input with incompatible shape (None, 3).\n", | |
| "prediction [[0.03814456]\n", | |
| " [0.9462323 ]\n", | |
| " [0.9463794 ]\n", | |
| " [0.05433232]]\n" | |
| ], | |
| "name": "stdout" | |
| } | |
| ] | |
| } | |
| ] | |
| } |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment