assignment1/softmax.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "6KY8i5AGAjsl",
    "tags": [
     "pdf-title"
    ]
   },
   "source": [
    "# Softmax exercise\n",
    "\n",
    "*Complete and hand in this completed worksheet (including its outputs and any supporting code outside of the worksheet) with your assignment submission. For more details see the [assignments page](http://vision.stanford.edu/teaching/cs231n/assignments.html) on the course website.*\n",
    "\n",
    "This exercise is analogous to the SVM exercise. You will:\n",
    "\n",
    "- implement a fully-vectorized **loss function** for the Softmax classifier\n",
    "- implement the fully-vectorized expression for its **analytic gradient**\n",
    "- **check your implementation** with numerical gradient\n",
    "- use a validation set to **tune the learning rate and regularization** strength\n",
    "- **optimize** the loss function with **SGD**\n",
    "- **visualize** the final learned weights\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "IU5ojYMuAjsp",
    "tags": [
     "pdf-ignore"
    ]
   },
   "outputs": [],
   "source": [
    "import random\n",
    "import numpy as np\n",
    "from cs231n.data_utils import load_CIFAR10\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "%matplotlib inline\n",
    "plt.rcParams['figure.figsize'] = (10.0, 8.0) # set default size of plots\n",
    "plt.rcParams['image.interpolation'] = 'nearest'\n",
    "plt.rcParams['image.cmap'] = 'gray'\n",
    "\n",
    "# for auto-reloading extenrnal modules\n",
    "# see http://stackoverflow.com/questions/1907993/autoreload-of-modules-in-ipython\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 163
    },
    "colab_type": "code",
    "id": "A59CP4hSAjs2",
    "outputId": "3e5e0045-513c-4341-d542-84ff5bc4bd93",
    "tags": [
     "pdf-ignore"
    ]
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train data shape:  (49000, 3073)\n",
      "Train labels shape:  (49000,)\n",
      "Validation data shape:  (1000, 3073)\n",
      "Validation labels shape:  (1000,)\n",
      "Test data shape:  (1000, 3073)\n",
      "Test labels shape:  (1000,)\n",
      "dev data shape:  (500, 3073)\n",
      "dev labels shape:  (500,)\n"
     ]
    }
   ],
   "source": [
    "def get_CIFAR10_data(num_training=49000, num_validation=1000, num_test=1000, num_dev=500):\n",
    "    \"\"\"\n",
    "    Load the CIFAR-10 dataset from disk and perform preprocessing to prepare\n",
    "    it for the linear classifier. These are the same steps as we used for the\n",
    "    SVM, but condensed to a single function.  \n",
    "    \"\"\"\n",
    "    # Load the raw CIFAR-10 data\n",
    "    cifar10_dir = 'cs231n/datasets/cifar-10-batches-py'\n",
    "    \n",
    "    # Cleaning up variables to prevent loading data multiple times (which may cause memory issue)\n",
    "    try:\n",
    "       del X_train, y_train\n",
    "       del X_test, y_test\n",
    "       print('Clear previously loaded data.')\n",
    "    except:\n",
    "       pass\n",
    "\n",
    "    X_train, y_train, X_test, y_test = load_CIFAR10(cifar10_dir)\n",
    "    \n",
    "    # subsample the data\n",
    "    mask = list(range(num_training, num_training + num_validation))\n",
    "    X_val = X_train[mask]\n",
    "    y_val = y_train[mask]\n",
    "    mask = list(range(num_training))\n",
    "    X_train = X_train[mask]\n",
    "    y_train = y_train[mask]\n",
    "    mask = list(range(num_test))\n",
    "    X_test = X_test[mask]\n",
    "    y_test = y_test[mask]\n",
    "    mask = np.random.choice(num_training, num_dev, replace=False)\n",
    "    X_dev = X_train[mask]\n",
    "    y_dev = y_train[mask]\n",
    "    \n",
    "    # Preprocessing: reshape the image data into rows\n",
    "    X_train = np.reshape(X_train, (X_train.shape[0], -1))\n",
    "    X_val = np.reshape(X_val, (X_val.shape[0], -1))\n",
    "    X_test = np.reshape(X_test, (X_test.shape[0], -1))\n",
    "    X_dev = np.reshape(X_dev, (X_dev.shape[0], -1))\n",
    "    \n",
    "    # Normalize the data: subtract the mean image\n",
    "    mean_image = np.mean(X_train, axis = 0)\n",
    "    X_train -= mean_image\n",
    "    X_val -= mean_image\n",
    "    X_test -= mean_image\n",
    "    X_dev -= mean_image\n",
    "    \n",
    "    # add bias dimension and transform into columns\n",
    "    X_train = np.hstack([X_train, np.ones((X_train.shape[0], 1))])\n",
    "    X_val = np.hstack([X_val, np.ones((X_val.shape[0], 1))])\n",
    "    X_test = np.hstack([X_test, np.ones((X_test.shape[0], 1))])\n",
    "    X_dev = np.hstack([X_dev, np.ones((X_dev.shape[0], 1))])\n",
    "    \n",
    "    return X_train, y_train, X_val, y_val, X_test, y_test, X_dev, y_dev\n",
    "\n",
    "\n",
    "# Invoke the above function to get our data.\n",
    "X_train, y_train, X_val, y_val, X_test, y_test, X_dev, y_dev = get_CIFAR10_data()\n",
    "print('Train data shape: ', X_train.shape)\n",
    "print('Train labels shape: ', y_train.shape)\n",
    "print('Validation data shape: ', X_val.shape)\n",
    "print('Validation labels shape: ', y_val.shape)\n",
    "print('Test data shape: ', X_test.shape)\n",
    "print('Test labels shape: ', y_test.shape)\n",
    "print('dev data shape: ', X_dev.shape)\n",
    "print('dev labels shape: ', y_dev.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "sT1LbMv6AjtC"
   },
   "source": [
    "## Softmax Classifier\n",
    "\n",
    "Your code for this section will all be written inside **cs231n/classifiers/softmax.py**. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 54
    },
    "colab_type": "code",
    "id": "mqJ7QsNdAjtG",
    "outputId": "2f0e7dc3-f966-439c-e2c4-14d377c866a1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loss: 2.343978\n",
      "sanity check: 2.302585\n"
     ]
    }
   ],
   "source": [
    "# First implement the naive softmax loss function with nested loops.\n",
    "# Open the file cs231n/classifiers/softmax.py and implement the\n",
    "# softmax_loss_naive function.\n",
    "\n",
    "from cs231n.classifiers.softmax import softmax_loss_naive\n",
    "import time\n",
    "\n",
    "# Generate a random softmax weight matrix and use it to compute the loss.\n",
    "W = np.random.randn(3073, 10) * 0.0001\n",
    "loss, grad = softmax_loss_naive(W, X_dev, y_dev, 0.0)\n",
    "\n",
    "# As a rough sanity check, our loss should be something close to -log(0.1).\n",
    "print('loss: %f' % loss)\n",
    "print('sanity check: %f' % (-np.log(0.1)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Pkxc2pyGAjtQ",
    "tags": [
     "pdf-inline"
    ]
   },
   "source": [
    "**Inline Question 1**\n",
    "\n",
    "Why do we expect our loss to be close to -log(0.1)? Explain briefly.**\n",
    "\n",
    "$\\color{blue}{\\textit Your Answer:}$ *When W is nearly a zero matrix loss = -log(1/number of classes). here number of classes=10 hence loss = -log(0.1)* \n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 381
    },
    "colab_type": "code",
    "id": "uR41k1QJAjtT",
    "outputId": "977004c0-c3cb-4697-fc69-b1c98d0dce4f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "numerical: 2.171142 analytic: 2.171141, relative error: 1.627266e-08\n",
      "numerical: 0.400820 analytic: 0.400820, relative error: 5.040518e-08\n",
      "numerical: 1.380600 analytic: 1.380600, relative error: 4.551984e-08\n",
      "numerical: -2.040608 analytic: -2.040608, relative error: 4.882013e-09\n",
      "numerical: 0.577150 analytic: 0.577150, relative error: 7.138072e-08\n",
      "numerical: -0.277016 analytic: -0.277016, relative error: 2.390908e-07\n",
      "numerical: -0.889613 analytic: -0.889613, relative error: 2.807468e-08\n",
      "numerical: -5.039652 analytic: -5.039652, relative error: 1.032463e-08\n",
      "numerical: -0.020561 analytic: -0.020561, relative error: 8.030967e-07\n",
      "numerical: -1.146366 analytic: -1.146366, relative error: 8.312084e-09\n",
      "numerical: -2.837108 analytic: -2.837108, relative error: 1.218638e-08\n",
      "numerical: 2.195394 analytic: 2.195394, relative error: 2.770568e-08\n",
      "numerical: 0.850851 analytic: 0.850851, relative error: 2.998686e-08\n",
      "numerical: 0.631710 analytic: 0.631710, relative error: 7.744426e-08\n",
      "numerical: -2.336302 analytic: -2.336302, relative error: 6.923597e-09\n",
      "numerical: 1.446800 analytic: 1.446800, relative error: 1.211929e-08\n",
      "numerical: 0.427467 analytic: 0.427467, relative error: 4.715276e-08\n",
      "numerical: 1.635089 analytic: 1.635089, relative error: 1.690012e-08\n",
      "numerical: 1.647170 analytic: 1.647169, relative error: 5.062822e-08\n",
      "numerical: -0.107483 analytic: -0.107483, relative error: 2.625195e-10\n"
     ]
    }
   ],
   "source": [
    "# Complete the implementation of softmax_loss_naive and implement a (naive)\n",
    "# version of the gradient that uses nested loops.\n",
    "loss, grad = softmax_loss_naive(W, X_dev, y_dev, 0.0)\n",
    "\n",
    "# As we did for the SVM, use numeric gradient checking as a debugging tool.\n",
    "# The numeric gradient should be close to the analytic gradient.\n",
    "from cs231n.gradient_check import grad_check_sparse\n",
    "f = lambda w: softmax_loss_naive(w, X_dev, y_dev, 0.0)[0]\n",
    "grad_numerical = grad_check_sparse(f, W, grad, 10)\n",
    "\n",
    "# similar to SVM case, do another gradient check with regularization\n",
    "loss, grad = softmax_loss_naive(W, X_dev, y_dev, 5e1)\n",
    "f = lambda w: softmax_loss_naive(w, X_dev, y_dev, 5e1)[0]\n",
    "grad_numerical = grad_check_sparse(f, W, grad, 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 90
    },
    "colab_type": "code",
    "id": "1S-xMiRvAjtb",
    "outputId": "2f90c456-f634-4d25-fa9d-7c14eae4d47f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "naive loss: 2.343978e+00 computed in 0.132757s\n",
      "vectorized loss: 2.343978e+00 computed in 0.007698s\n",
      "Loss difference: 0.000000\n",
      "Gradient difference: 0.000000\n"
     ]
    }
   ],
   "source": [
    "# Now that we have a naive implementation of the softmax loss function and its gradient,\n",
    "# implement a vectorized version in softmax_loss_vectorized.\n",
    "# The two versions should compute the same results, but the vectorized version should be\n",
    "# much faster.\n",
    "tic = time.time()\n",
    "loss_naive, grad_naive = softmax_loss_naive(W, X_dev, y_dev, 0.000005)\n",
    "toc = time.time()\n",
    "print('naive loss: %e computed in %fs' % (loss_naive, toc - tic))\n",
    "\n",
    "from cs231n.classifiers.softmax import softmax_loss_vectorized\n",
    "tic = time.time()\n",
    "loss_vectorized, grad_vectorized = softmax_loss_vectorized(W, X_dev, y_dev, 0.000005)\n",
    "toc = time.time()\n",
    "print('vectorized loss: %e computed in %fs' % (loss_vectorized, toc - tic))\n",
    "\n",
    "# As we did for the SVM, we use the Frobenius norm to compare the two versions\n",
    "# of the gradient.\n",
    "grad_difference = np.linalg.norm(grad_naive - grad_vectorized, ord='fro')\n",
    "print('Loss difference: %f' % np.abs(loss_naive - loss_vectorized))\n",
    "print('Gradient difference: %f' % grad_difference)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 8763
    },
    "colab_type": "code",
    "id": "--MckoWYAjtm",
    "outputId": "d010a2eb-d91b-4a66-99d6-ab37370e3f84",
    "tags": [
     "code"
    ]
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration 0 / 1500: loss 310.617131\n",
      "iteration 100 / 1500: loss 224.595166\n",
      "iteration 200 / 1500: loss 163.056724\n",
      "iteration 300 / 1500: loss 118.569695\n",
      "iteration 400 / 1500: loss 86.441031\n",
      "iteration 500 / 1500: loss 63.270467\n",
      "iteration 600 / 1500: loss 46.384703\n",
      "iteration 700 / 1500: loss 34.207405\n",
      "iteration 800 / 1500: loss 25.355558\n",
      "iteration 900 / 1500: loss 18.886004\n",
      "iteration 1000 / 1500: loss 14.143174\n",
      "iteration 1100 / 1500: loss 10.910826\n",
      "iteration 1200 / 1500: loss 8.515440\n",
      "iteration 1300 / 1500: loss 6.623831\n",
      "iteration 1400 / 1500: loss 5.417502\n",
      "iteration 0 / 1500: loss 935.921905\n",
      "iteration 100 / 1500: loss 357.629589\n",
      "iteration 200 / 1500: loss 137.716507\n",
      "iteration 300 / 1500: loss 53.817592\n",
      "iteration 400 / 1500: loss 21.784637\n",
      "iteration 500 / 1500: loss 9.591403\n",
      "iteration 600 / 1500: loss 4.971029\n",
      "iteration 700 / 1500: loss 3.182322\n",
      "iteration 800 / 1500: loss 2.504187\n",
      "iteration 900 / 1500: loss 2.281046\n",
      "iteration 1000 / 1500: loss 2.192228\n",
      "iteration 1100 / 1500: loss 2.126766\n",
      "iteration 1200 / 1500: loss 2.028538\n",
      "iteration 1300 / 1500: loss 2.137832\n",
      "iteration 1400 / 1500: loss 2.082869\n",
      "iteration 0 / 1500: loss 1554.697742\n",
      "iteration 100 / 1500: loss 312.594295\n",
      "iteration 200 / 1500: loss 64.272036\n",
      "iteration 300 / 1500: loss 14.577173\n",
      "iteration 400 / 1500: loss 4.646553\n",
      "iteration 500 / 1500: loss 2.663291\n",
      "iteration 600 / 1500: loss 2.215335\n",
      "iteration 700 / 1500: loss 2.146672\n",
      "iteration 800 / 1500: loss 2.141930\n",
      "iteration 900 / 1500: loss 2.189370\n",
      "iteration 1000 / 1500: loss 2.141948\n",
      "iteration 1100 / 1500: loss 2.212168\n",
      "iteration 1200 / 1500: loss 2.143262\n",
      "iteration 1300 / 1500: loss 2.117046\n",
      "iteration 1400 / 1500: loss 2.106850\n",
      "iteration 0 / 1500: loss 311.372715\n",
      "iteration 100 / 1500: loss 213.557320\n",
      "iteration 200 / 1500: loss 147.369584\n",
      "iteration 300 / 1500: loss 101.856366\n",
      "iteration 400 / 1500: loss 70.520567\n",
      "iteration 500 / 1500: loss 48.943753\n",
      "iteration 600 / 1500: loss 34.304814\n",
      "iteration 700 / 1500: loss 24.244155\n",
      "iteration 800 / 1500: loss 17.185694\n",
      "iteration 900 / 1500: loss 12.518254\n",
      "iteration 1000 / 1500: loss 9.207089\n",
      "iteration 1100 / 1500: loss 6.942207\n",
      "iteration 1200 / 1500: loss 5.377445\n",
      "iteration 1300 / 1500: loss 4.305663\n",
      "iteration 1400 / 1500: loss 3.557628\n",
      "iteration 0 / 1500: loss 930.532266\n",
      "iteration 100 / 1500: loss 303.215150\n",
      "iteration 200 / 1500: loss 99.766567\n",
      "iteration 300 / 1500: loss 33.772097\n",
      "iteration 400 / 1500: loss 12.451423\n",
      "iteration 500 / 1500: loss 5.419028\n",
      "iteration 600 / 1500: loss 3.214060\n",
      "iteration 700 / 1500: loss 2.467572\n",
      "iteration 800 / 1500: loss 2.199809\n",
      "iteration 900 / 1500: loss 2.157359\n",
      "iteration 1000 / 1500: loss 2.083930\n",
      "iteration 1100 / 1500: loss 2.119768\n",
      "iteration 1200 / 1500: loss 2.109730\n",
      "iteration 1300 / 1500: loss 2.141394\n",
      "iteration 1400 / 1500: loss 2.104656\n",
      "iteration 0 / 1500: loss 1548.618466\n",
      "iteration 100 / 1500: loss 238.103447\n",
      "iteration 200 / 1500: loss 38.181273\n",
      "iteration 300 / 1500: loss 7.669166\n",
      "iteration 400 / 1500: loss 2.977432\n",
      "iteration 500 / 1500: loss 2.277126\n",
      "iteration 600 / 1500: loss 2.151349\n",
      "iteration 700 / 1500: loss 2.146350\n",
      "iteration 800 / 1500: loss 2.115745\n",
      "iteration 900 / 1500: loss 2.128591\n",
      "iteration 1000 / 1500: loss 2.169853\n",
      "iteration 1100 / 1500: loss 2.140476\n",
      "iteration 1200 / 1500: loss 2.142624\n",
      "iteration 1300 / 1500: loss 2.178460\n",
      "iteration 1400 / 1500: loss 2.145708\n",
      "iteration 0 / 1500: loss 312.367522\n",
      "iteration 100 / 1500: loss 202.550051\n",
      "iteration 200 / 1500: loss 132.583062\n",
      "iteration 300 / 1500: loss 86.886728\n",
      "iteration 400 / 1500: loss 57.146204\n",
      "iteration 500 / 1500: loss 38.028880\n",
      "iteration 600 / 1500: loss 25.418424\n",
      "iteration 700 / 1500: loss 17.217708\n",
      "iteration 800 / 1500: loss 11.927329\n",
      "iteration 900 / 1500: loss 8.453804\n",
      "iteration 1000 / 1500: loss 6.279640\n",
      "iteration 1100 / 1500: loss 4.711325\n",
      "iteration 1200 / 1500: loss 3.786666\n",
      "iteration 1300 / 1500: loss 3.192932\n",
      "iteration 1400 / 1500: loss 2.710198\n",
      "iteration 0 / 1500: loss 922.166309\n",
      "iteration 100 / 1500: loss 255.946097\n",
      "iteration 200 / 1500: loss 72.105442\n",
      "iteration 300 / 1500: loss 21.464474\n",
      "iteration 400 / 1500: loss 7.388247\n",
      "iteration 500 / 1500: loss 3.568238\n",
      "iteration 600 / 1500: loss 2.519054\n",
      "iteration 700 / 1500: loss 2.215859\n",
      "iteration 800 / 1500: loss 2.151468\n",
      "iteration 900 / 1500: loss 2.147367\n",
      "iteration 1000 / 1500: loss 2.134104\n",
      "iteration 1100 / 1500: loss 2.105823\n",
      "iteration 1200 / 1500: loss 2.092848\n",
      "iteration 1300 / 1500: loss 2.103473\n",
      "iteration 1400 / 1500: loss 2.106784\n",
      "iteration 0 / 1500: loss 1543.071596\n",
      "iteration 100 / 1500: loss 181.939454\n",
      "iteration 200 / 1500: loss 23.085000\n",
      "iteration 300 / 1500: loss 4.569042\n",
      "iteration 400 / 1500: loss 2.435229\n",
      "iteration 500 / 1500: loss 2.192847\n",
      "iteration 600 / 1500: loss 2.138415\n",
      "iteration 700 / 1500: loss 2.083169\n",
      "iteration 800 / 1500: loss 2.122180\n",
      "iteration 900 / 1500: loss 2.150655\n",
      "iteration 1000 / 1500: loss 2.163679\n",
      "iteration 1100 / 1500: loss 2.132682\n",
      "iteration 1200 / 1500: loss 2.137430\n",
      "iteration 1300 / 1500: loss 2.150367\n",
      "iteration 1400 / 1500: loss 2.153142\n",
      "iteration 0 / 1500: loss 313.704944\n",
      "iteration 100 / 1500: loss 193.116923\n",
      "iteration 200 / 1500: loss 119.738763\n",
      "iteration 300 / 1500: loss 74.611348\n",
      "iteration 400 / 1500: loss 46.708053\n",
      "iteration 500 / 1500: loss 29.562661\n",
      "iteration 600 / 1500: loss 19.012238\n",
      "iteration 700 / 1500: loss 12.513927\n",
      "iteration 800 / 1500: loss 8.483225\n",
      "iteration 900 / 1500: loss 6.042146\n",
      "iteration 1000 / 1500: loss 4.416312\n",
      "iteration 1100 / 1500: loss 3.478122\n",
      "iteration 1200 / 1500: loss 2.907312\n",
      "iteration 1300 / 1500: loss 2.538820\n",
      "iteration 1400 / 1500: loss 2.304153\n",
      "iteration 0 / 1500: loss 934.596915\n",
      "iteration 100 / 1500: loss 220.626502\n",
      "iteration 200 / 1500: loss 53.401986\n",
      "iteration 300 / 1500: loss 14.108573\n",
      "iteration 400 / 1500: loss 4.937080\n",
      "iteration 500 / 1500: loss 2.731236\n",
      "iteration 600 / 1500: loss 2.267032\n",
      "iteration 700 / 1500: loss 2.125392\n",
      "iteration 800 / 1500: loss 2.105019\n",
      "iteration 900 / 1500: loss 2.008848\n",
      "iteration 1000 / 1500: loss 2.153674\n",
      "iteration 1100 / 1500: loss 2.089494\n",
      "iteration 1200 / 1500: loss 2.074332\n",
      "iteration 1300 / 1500: loss 2.074691\n",
      "iteration 1400 / 1500: loss 2.064070\n",
      "iteration 0 / 1500: loss 1541.993535\n",
      "iteration 100 / 1500: loss 139.259914\n",
      "iteration 200 / 1500: loss 14.336848\n",
      "iteration 300 / 1500: loss 3.232654\n",
      "iteration 400 / 1500: loss 2.258844\n",
      "iteration 500 / 1500: loss 2.107965\n",
      "iteration 600 / 1500: loss 2.147723\n",
      "iteration 700 / 1500: loss 2.072657\n",
      "iteration 800 / 1500: loss 2.174096\n",
      "iteration 900 / 1500: loss 2.169460\n",
      "iteration 1000 / 1500: loss 2.120635\n",
      "iteration 1100 / 1500: loss 2.082734\n",
      "iteration 1200 / 1500: loss 2.151530\n",
      "iteration 1300 / 1500: loss 2.128975\n",
      "iteration 1400 / 1500: loss 2.165589\n",
      "iteration 0 / 1500: loss 310.604929\n",
      "iteration 100 / 1500: loss 181.422983\n",
      "iteration 200 / 1500: loss 106.757244\n",
      "iteration 300 / 1500: loss 63.079376\n",
      "iteration 400 / 1500: loss 37.816055\n",
      "iteration 500 / 1500: loss 22.846148\n",
      "iteration 600 / 1500: loss 14.328654\n",
      "iteration 700 / 1500: loss 9.125902\n",
      "iteration 800 / 1500: loss 6.210727\n",
      "iteration 900 / 1500: loss 4.476588\n",
      "iteration 1000 / 1500: loss 3.400289\n",
      "iteration 1100 / 1500: loss 2.808420\n",
      "iteration 1200 / 1500: loss 2.465372\n",
      "iteration 1300 / 1500: loss 2.364860\n",
      "iteration 1400 / 1500: loss 2.145340\n",
      "iteration 0 / 1500: loss 922.463394\n",
      "iteration 100 / 1500: loss 185.683734\n",
      "iteration 200 / 1500: loss 38.862814\n",
      "iteration 300 / 1500: loss 9.445246\n",
      "iteration 400 / 1500: loss 3.578688\n",
      "iteration 500 / 1500: loss 2.304395\n",
      "iteration 600 / 1500: loss 2.134142\n",
      "iteration 700 / 1500: loss 2.117295\n",
      "iteration 800 / 1500: loss 2.118852\n",
      "iteration 900 / 1500: loss 2.107571\n",
      "iteration 1000 / 1500: loss 2.108083\n",
      "iteration 1100 / 1500: loss 2.099707\n",
      "iteration 1200 / 1500: loss 2.119551\n",
      "iteration 1300 / 1500: loss 2.037194\n",
      "iteration 1400 / 1500: loss 2.067174\n",
      "iteration 0 / 1500: loss 1540.935690\n",
      "iteration 100 / 1500: loss 106.595022\n",
      "iteration 200 / 1500: loss 9.240530\n",
      "iteration 300 / 1500: loss 2.635304\n",
      "iteration 400 / 1500: loss 2.127572\n",
      "iteration 500 / 1500: loss 2.120837\n",
      "iteration 600 / 1500: loss 2.182760\n",
      "iteration 700 / 1500: loss 2.184524\n",
      "iteration 800 / 1500: loss 2.137389\n",
      "iteration 900 / 1500: loss 2.139310\n",
      "iteration 1000 / 1500: loss 2.152934\n",
      "iteration 1100 / 1500: loss 2.149676\n",
      "iteration 1200 / 1500: loss 2.165373\n",
      "iteration 1300 / 1500: loss 2.124982\n",
      "iteration 1400 / 1500: loss 2.081777\n",
      "iteration 0 / 1500: loss 315.123960\n",
      "iteration 100 / 1500: loss 174.695029\n",
      "iteration 200 / 1500: loss 97.473334\n",
      "iteration 300 / 1500: loss 54.889482\n",
      "iteration 400 / 1500: loss 31.392442\n",
      "iteration 500 / 1500: loss 18.260533\n",
      "iteration 600 / 1500: loss 10.997835\n",
      "iteration 700 / 1500: loss 7.036047\n",
      "iteration 800 / 1500: loss 4.754127\n",
      "iteration 900 / 1500: loss 3.626011\n",
      "iteration 1000 / 1500: loss 2.827412\n",
      "iteration 1100 / 1500: loss 2.463293\n",
      "iteration 1200 / 1500: loss 2.244194\n",
      "iteration 1300 / 1500: loss 2.134775\n",
      "iteration 1400 / 1500: loss 2.032935\n",
      "iteration 0 / 1500: loss 916.000609\n",
      "iteration 100 / 1500: loss 157.243259\n",
      "iteration 200 / 1500: loss 28.388875\n",
      "iteration 300 / 1500: loss 6.581543\n",
      "iteration 400 / 1500: loss 2.858438\n",
      "iteration 500 / 1500: loss 2.260983\n",
      "iteration 600 / 1500: loss 2.156522\n",
      "iteration 700 / 1500: loss 2.041364\n",
      "iteration 800 / 1500: loss 2.116799\n",
      "iteration 900 / 1500: loss 2.109694\n",
      "iteration 1000 / 1500: loss 2.089272\n",
      "iteration 1100 / 1500: loss 2.135500\n",
      "iteration 1200 / 1500: loss 2.093520\n",
      "iteration 1300 / 1500: loss 2.102514\n",
      "iteration 1400 / 1500: loss 2.030809\n",
      "iteration 0 / 1500: loss 1549.544313\n",
      "iteration 100 / 1500: loss 82.255890\n",
      "iteration 200 / 1500: loss 6.264471\n",
      "iteration 300 / 1500: loss 2.375126\n",
      "iteration 400 / 1500: loss 2.075523\n",
      "iteration 500 / 1500: loss 2.095585\n",
      "iteration 600 / 1500: loss 2.131714\n",
      "iteration 700 / 1500: loss 2.091467\n",
      "iteration 800 / 1500: loss 2.163182\n",
      "iteration 900 / 1500: loss 2.097679\n",
      "iteration 1000 / 1500: loss 2.176837\n",
      "iteration 1100 / 1500: loss 2.155797\n",
      "iteration 1200 / 1500: loss 2.176184\n",
      "iteration 1300 / 1500: loss 2.177547\n",
      "iteration 1400 / 1500: loss 2.160146\n",
      "iteration 0 / 1500: loss 314.206212\n",
      "iteration 100 / 1500: loss 165.134145\n",
      "iteration 200 / 1500: loss 87.494293\n",
      "iteration 300 / 1500: loss 46.869422\n",
      "iteration 400 / 1500: loss 25.595540\n",
      "iteration 500 / 1500: loss 14.400914\n",
      "iteration 600 / 1500: loss 8.468887\n",
      "iteration 700 / 1500: loss 5.440016\n",
      "iteration 800 / 1500: loss 3.840071\n",
      "iteration 900 / 1500: loss 2.932498\n",
      "iteration 1000 / 1500: loss 2.470938\n",
      "iteration 1100 / 1500: loss 2.251807\n",
      "iteration 1200 / 1500: loss 2.137809\n",
      "iteration 1300 / 1500: loss 2.112261\n",
      "iteration 1400 / 1500: loss 2.012426\n",
      "iteration 0 / 1500: loss 924.748987\n",
      "iteration 100 / 1500: loss 135.337881\n",
      "iteration 200 / 1500: loss 21.343051\n",
      "iteration 300 / 1500: loss 4.886975\n",
      "iteration 400 / 1500: loss 2.525873\n",
      "iteration 500 / 1500: loss 2.185811\n",
      "iteration 600 / 1500: loss 2.100544\n",
      "iteration 700 / 1500: loss 2.096842\n",
      "iteration 800 / 1500: loss 2.099657\n",
      "iteration 900 / 1500: loss 2.131708\n",
      "iteration 1000 / 1500: loss 2.053307\n",
      "iteration 1100 / 1500: loss 2.097689\n",
      "iteration 1200 / 1500: loss 2.071473\n",
      "iteration 1300 / 1500: loss 2.108229\n",
      "iteration 1400 / 1500: loss 2.131179\n",
      "iteration 0 / 1500: loss 1534.251426\n",
      "iteration 100 / 1500: loss 62.623880\n",
      "iteration 200 / 1500: loss 4.538014\n",
      "iteration 300 / 1500: loss 2.240452\n",
      "iteration 400 / 1500: loss 2.177715\n",
      "iteration 500 / 1500: loss 2.158280\n",
      "iteration 600 / 1500: loss 2.198116\n",
      "iteration 700 / 1500: loss 2.173936\n",
      "iteration 800 / 1500: loss 2.159473\n",
      "iteration 900 / 1500: loss 2.165116\n",
      "iteration 1000 / 1500: loss 2.136442\n",
      "iteration 1100 / 1500: loss 2.183531\n",
      "iteration 1200 / 1500: loss 2.164129\n",
      "iteration 1300 / 1500: loss 2.113189\n",
      "iteration 1400 / 1500: loss 2.129227\n",
      "iteration 0 / 1500: loss 317.165146\n",
      "iteration 100 / 1500: loss 158.254970\n",
      "iteration 200 / 1500: loss 79.541913\n",
      "iteration 300 / 1500: loss 40.540486\n",
      "iteration 400 / 1500: loss 21.315517\n",
      "iteration 500 / 1500: loss 11.592421\n",
      "iteration 600 / 1500: loss 6.777641\n",
      "iteration 700 / 1500: loss 4.382143\n",
      "iteration 800 / 1500: loss 3.194647\n",
      "iteration 900 / 1500: loss 2.571585\n",
      "iteration 1000 / 1500: loss 2.280124\n",
      "iteration 1100 / 1500: loss 2.142294\n",
      "iteration 1200 / 1500: loss 2.020087\n",
      "iteration 1300 / 1500: loss 2.043854\n",
      "iteration 1400 / 1500: loss 2.014351\n",
      "iteration 0 / 1500: loss 930.435277\n",
      "iteration 100 / 1500: loss 116.088340\n",
      "iteration 200 / 1500: loss 16.203170\n",
      "iteration 300 / 1500: loss 3.806013\n",
      "iteration 400 / 1500: loss 2.339013\n",
      "iteration 500 / 1500: loss 2.138658\n",
      "iteration 600 / 1500: loss 2.054785\n",
      "iteration 700 / 1500: loss 1.999515\n",
      "iteration 800 / 1500: loss 2.136463\n",
      "iteration 900 / 1500: loss 2.123994\n",
      "iteration 1000 / 1500: loss 2.106332\n",
      "iteration 1100 / 1500: loss 2.057160\n",
      "iteration 1200 / 1500: loss 2.125186\n",
      "iteration 1300 / 1500: loss 2.110090\n",
      "iteration 1400 / 1500: loss 2.111083\n",
      "iteration 0 / 1500: loss 1541.048307\n",
      "iteration 100 / 1500: loss 48.450848\n",
      "iteration 200 / 1500: loss 3.518315\n",
      "iteration 300 / 1500: loss 2.194309\n",
      "iteration 400 / 1500: loss 2.131435\n",
      "iteration 500 / 1500: loss 2.190628\n",
      "iteration 600 / 1500: loss 2.074024\n",
      "iteration 700 / 1500: loss 2.161487\n",
      "iteration 800 / 1500: loss 2.167155\n",
      "iteration 900 / 1500: loss 2.154900\n",
      "iteration 1000 / 1500: loss 2.135844\n",
      "iteration 1100 / 1500: loss 2.164268\n",
      "iteration 1200 / 1500: loss 2.193256\n",
      "iteration 1300 / 1500: loss 2.166910\n",
      "iteration 1400 / 1500: loss 2.122904\n",
      "iteration 0 / 1500: loss 311.763777\n",
      "iteration 100 / 1500: loss 147.445604\n",
      "iteration 200 / 1500: loss 70.564022\n",
      "iteration 300 / 1500: loss 34.197546\n",
      "iteration 400 / 1500: loss 17.327608\n",
      "iteration 500 / 1500: loss 9.189986\n",
      "iteration 600 / 1500: loss 5.403852\n",
      "iteration 700 / 1500: loss 3.587124\n",
      "iteration 800 / 1500: loss 2.743299\n",
      "iteration 900 / 1500: loss 2.271043\n",
      "iteration 1000 / 1500: loss 2.142631\n",
      "iteration 1100 / 1500: loss 2.095474\n",
      "iteration 1200 / 1500: loss 1.941218\n",
      "iteration 1300 / 1500: loss 2.031264\n",
      "iteration 1400 / 1500: loss 2.016865\n",
      "iteration 0 / 1500: loss 924.801486\n",
      "iteration 100 / 1500: loss 98.366758\n",
      "iteration 200 / 1500: loss 12.183586\n",
      "iteration 300 / 1500: loss 3.142507\n",
      "iteration 400 / 1500: loss 2.190644\n",
      "iteration 500 / 1500: loss 2.115459\n",
      "iteration 600 / 1500: loss 2.106661\n",
      "iteration 700 / 1500: loss 2.076808\n",
      "iteration 800 / 1500: loss 2.104784\n",
      "iteration 900 / 1500: loss 2.108668\n",
      "iteration 1000 / 1500: loss 2.122773\n",
      "iteration 1100 / 1500: loss 2.109224\n",
      "iteration 1200 / 1500: loss 2.135413\n",
      "iteration 1300 / 1500: loss 2.131743\n",
      "iteration 1400 / 1500: loss 2.087213\n",
      "iteration 0 / 1500: loss 1543.589724\n",
      "iteration 100 / 1500: loss 37.507159\n",
      "iteration 200 / 1500: loss 2.952010\n",
      "iteration 300 / 1500: loss 2.160092\n",
      "iteration 400 / 1500: loss 2.168209\n",
      "iteration 500 / 1500: loss 2.185058\n",
      "iteration 600 / 1500: loss 2.178453\n",
      "iteration 700 / 1500: loss 2.146778\n",
      "iteration 800 / 1500: loss 2.116545\n",
      "iteration 900 / 1500: loss 2.139824\n",
      "iteration 1000 / 1500: loss 2.159814\n",
      "iteration 1100 / 1500: loss 2.173429\n",
      "iteration 1200 / 1500: loss 2.122218\n",
      "iteration 1300 / 1500: loss 2.193761\n",
      "iteration 1400 / 1500: loss 2.173293\n",
      "iteration 0 / 1500: loss 313.402361\n",
      "iteration 100 / 1500: loss 140.536142\n",
      "iteration 200 / 1500: loss 63.682898\n",
      "iteration 300 / 1500: loss 29.611825\n",
      "iteration 400 / 1500: loss 14.301333\n",
      "iteration 500 / 1500: loss 7.466559\n",
      "iteration 600 / 1500: loss 4.538265\n",
      "iteration 700 / 1500: loss 3.167983\n",
      "iteration 800 / 1500: loss 2.499225\n",
      "iteration 900 / 1500: loss 2.265847\n",
      "iteration 1000 / 1500: loss 2.131373\n",
      "iteration 1100 / 1500: loss 2.055394\n",
      "iteration 1200 / 1500: loss 2.046569\n",
      "iteration 1300 / 1500: loss 2.054130\n",
      "iteration 1400 / 1500: loss 2.055762\n",
      "iteration 0 / 1500: loss 939.518723\n",
      "iteration 100 / 1500: loss 85.313557\n",
      "iteration 200 / 1500: loss 9.561172\n",
      "iteration 300 / 1500: loss 2.783075\n",
      "iteration 400 / 1500: loss 2.159631\n",
      "iteration 500 / 1500: loss 2.061089\n",
      "iteration 600 / 1500: loss 2.046139\n",
      "iteration 700 / 1500: loss 2.095141\n",
      "iteration 800 / 1500: loss 2.064109\n",
      "iteration 900 / 1500: loss 2.049550\n",
      "iteration 1000 / 1500: loss 2.140778\n",
      "iteration 1100 / 1500: loss 2.149230\n",
      "iteration 1200 / 1500: loss 2.050340\n",
      "iteration 1300 / 1500: loss 2.153931\n",
      "iteration 1400 / 1500: loss 2.071488\n",
      "iteration 0 / 1500: loss 1525.719751\n",
      "iteration 100 / 1500: loss 28.750226\n",
      "iteration 200 / 1500: loss 2.612200\n",
      "iteration 300 / 1500: loss 2.146391\n",
      "iteration 400 / 1500: loss 2.107185\n",
      "iteration 500 / 1500: loss 2.134537\n",
      "iteration 600 / 1500: loss 2.151038\n",
      "iteration 700 / 1500: loss 2.137029\n",
      "iteration 800 / 1500: loss 2.169043\n",
      "iteration 900 / 1500: loss 2.162368\n",
      "iteration 1000 / 1500: loss 2.179183\n",
      "iteration 1100 / 1500: loss 2.191193\n",
      "iteration 1200 / 1500: loss 2.171608\n",
      "iteration 1300 / 1500: loss 2.149073\n",
      "iteration 1400 / 1500: loss 2.140208\n",
      "lr 8.000000e-08 reg 1.000000e+04 train accuracy: 0.345347 val accuracy: 0.362000\n",
      "lr 8.000000e-08 reg 3.000000e+04 train accuracy: 0.319714 val accuracy: 0.335000\n",
      "lr 8.000000e-08 reg 5.000000e+04 train accuracy: 0.301776 val accuracy: 0.317000\n",
      "lr 9.333333e-08 reg 1.000000e+04 train accuracy: 0.351673 val accuracy: 0.356000\n",
      "lr 9.333333e-08 reg 3.000000e+04 train accuracy: 0.319367 val accuracy: 0.334000\n",
      "lr 9.333333e-08 reg 5.000000e+04 train accuracy: 0.306265 val accuracy: 0.323000\n",
      "lr 1.066667e-07 reg 1.000000e+04 train accuracy: 0.355245 val accuracy: 0.373000\n",
      "lr 1.066667e-07 reg 3.000000e+04 train accuracy: 0.327959 val accuracy: 0.337000\n",
      "lr 1.066667e-07 reg 5.000000e+04 train accuracy: 0.314306 val accuracy: 0.327000\n",
      "lr 1.200000e-07 reg 1.000000e+04 train accuracy: 0.358755 val accuracy: 0.370000\n",
      "lr 1.200000e-07 reg 3.000000e+04 train accuracy: 0.319959 val accuracy: 0.326000\n",
      "lr 1.200000e-07 reg 5.000000e+04 train accuracy: 0.310714 val accuracy: 0.327000\n",
      "lr 1.333333e-07 reg 1.000000e+04 train accuracy: 0.357367 val accuracy: 0.378000\n",
      "lr 1.333333e-07 reg 3.000000e+04 train accuracy: 0.318143 val accuracy: 0.341000\n",
      "lr 1.333333e-07 reg 5.000000e+04 train accuracy: 0.317612 val accuracy: 0.332000\n",
      "lr 1.466667e-07 reg 1.000000e+04 train accuracy: 0.354612 val accuracy: 0.365000\n",
      "lr 1.466667e-07 reg 3.000000e+04 train accuracy: 0.329184 val accuracy: 0.336000\n",
      "lr 1.466667e-07 reg 5.000000e+04 train accuracy: 0.304510 val accuracy: 0.319000\n",
      "lr 1.600000e-07 reg 1.000000e+04 train accuracy: 0.354000 val accuracy: 0.377000\n",
      "lr 1.600000e-07 reg 3.000000e+04 train accuracy: 0.323571 val accuracy: 0.326000\n",
      "lr 1.600000e-07 reg 5.000000e+04 train accuracy: 0.309061 val accuracy: 0.329000\n",
      "lr 1.733333e-07 reg 1.000000e+04 train accuracy: 0.355796 val accuracy: 0.369000\n",
      "lr 1.733333e-07 reg 3.000000e+04 train accuracy: 0.315592 val accuracy: 0.330000\n",
      "lr 1.733333e-07 reg 5.000000e+04 train accuracy: 0.309367 val accuracy: 0.322000\n",
      "lr 1.866667e-07 reg 1.000000e+04 train accuracy: 0.355449 val accuracy: 0.360000\n",
      "lr 1.866667e-07 reg 3.000000e+04 train accuracy: 0.324000 val accuracy: 0.341000\n",
      "lr 1.866667e-07 reg 5.000000e+04 train accuracy: 0.308571 val accuracy: 0.320000\n",
      "lr 2.000000e-07 reg 1.000000e+04 train accuracy: 0.357673 val accuracy: 0.371000\n",
      "lr 2.000000e-07 reg 3.000000e+04 train accuracy: 0.321959 val accuracy: 0.340000\n",
      "lr 2.000000e-07 reg 5.000000e+04 train accuracy: 0.307163 val accuracy: 0.321000\n",
      "best validation accuracy achieved during cross-validation: 0.378000\n"
     ]
    }
   ],
   "source": [
    "# Use the validation set to tune hyperparameters (regularization strength and\n",
    "# learning rate). You should experiment with different ranges for the learning\n",
    "# rates and regularization strengths; if you are careful you should be able to\n",
    "# get a classification accuracy of over 0.35 on the validation set.\n",
    "from cs231n.classifiers import Softmax\n",
    "results = {}\n",
    "best_val = -1\n",
    "best_softmax = None\n",
    "learning_rates = np.linspace(8e-8, 2e-7,10)\n",
    "regularization_strengths = np.linspace(1e4, 5e4,3)\n",
    "\n",
    "################################################################################\n",
    "# TODO:                                                                        #\n",
    "# Use the validation set to set the learning rate and regularization strength. #\n",
    "# This should be identical to the validation that you did for the SVM; save    #\n",
    "# the best trained softmax classifer in best_softmax.                          #\n",
    "################################################################################\n",
    "# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****\n",
    "\n",
    "for lr in learning_rates:\n",
    "  for reg in regularization_strengths:\n",
    "    clf=Softmax()\n",
    "    loss_hist = clf.train(X_train, y_train, learning_rate=lr, reg=reg,\n",
    "                      num_iters=1500, verbose=True)\n",
    "    y_train_pred = clf.predict(X_train)\n",
    "    y_val_pred = clf.predict(X_val)\n",
    "    train_acc=np.mean(y_train == y_train_pred)\n",
    "    val_acc=np.mean(y_val == y_val_pred)\n",
    "    results[(lr,reg)]=train_acc,val_acc\n",
    "    if(val_acc>best_val):\n",
    "        best_softmax=clf\n",
    "        best_val=val_acc\n",
    "# *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****\n",
    "    \n",
    "# Print out results.\n",
    "for lr, reg in sorted(results):\n",
    "    train_accuracy, val_accuracy = results[(lr, reg)]\n",
    "    print('lr %e reg %e train accuracy: %f val accuracy: %f' % (\n",
    "                lr, reg, train_accuracy, val_accuracy))\n",
    "    \n",
    "print('best validation accuracy achieved during cross-validation: %f' % best_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "hqhVxnB4Ajtr",
    "outputId": "91d9ceb7-545a-4f9c-a283-698748c579be"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "softmax on raw pixels final test set accuracy: 0.371000\n"
     ]
    }
   ],
   "source": [
    "# evaluate on test set\n",
    "# Evaluate the best softmax on test set\n",
    "y_test_pred = best_softmax.predict(X_test)\n",
    "test_accuracy = np.mean(y_test == y_test_pred)\n",
    "print('softmax on raw pixels final test set accuracy: %f' % (test_accuracy, ))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "yl2tVmY8Ajtx",
    "tags": [
     "pdf-inline"
    ]
   },
   "source": [
    "**Inline Question 2** - *True or False*\n",
    "\n",
    "Suppose the overall training loss is defined as the sum of the per-datapoint loss over all training examples. It is possible to add a new datapoint to a training set that would leave the SVM loss unchanged, but this is not the case with the Softmax classifier loss.\n",
    "\n",
    "$\\color{blue}{\\textit Your Answer:}$True\n",
    "\n",
    "\n",
    "$\\color{blue}{\\textit Your Explanation:}$ In svm loss depends on the relative score and if the score of the added datapoint is 1(margin) less than the correct score then the total loss remains unchanged whereas in softmax every datapoint has some contribution in the total loss.\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 397
    },
    "colab_type": "code",
    "id": "ytc6GAiSAjt1",
    "outputId": "d63d89f8-1bcd-4e18-a79f-1d2f47542b37"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlAAAAF8CAYAAAAAZIWVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEARBECrk/wDOPdOviBfErwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x576 with 10 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the learned weights for each class\n",
    "w = best_softmax.W[:-1,:] # strip out the bias\n",
    "w = w.reshape(32, 32, 3, 10)\n",
    "\n",
    "w_min, w_max = np.min(w), np.max(w)\n",
    "\n",
    "classes = ['plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']\n",
    "for i in range(10):\n",
    "    plt.subplot(2, 5, i + 1)\n",
    "    \n",
    "    # Rescale the weights to be between 0 and 255\n",
    "    wimg = 255.0 * (w[:, :, :, i].squeeze() - w_min) / (w_max - w_min)\n",
    "    plt.imshow(wimg.astype('uint8'))\n",
    "    plt.axis('off')\n",
    "    plt.title(classes[i])"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "softmax.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}