13-mnist-without-keras-tf.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "source": [
        "from keras.datasets import mnist\n",
        "from keras.utils import to_categorical\n",
        "import numpy as np\n",
        "import math\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# load data\n",
        "(train_x, train_y), (test_x, test_y) = mnist.load_data()\n",
        "\n",
        "# split data\n",
        "num_data = 500\n",
        "train_x = train_x[:num_data]\n",
        "train_y = train_y[:num_data]\n",
        "\n",
        "\n",
        "# preprocess (one hot encoding)\n",
        "np.set_printoptions(linewidth=200)\n",
        "print('Before preprocess x \\n', np.array(train_x[0]), '\\n')\n",
        "print('Before preprocess y \\n', np.array(train_y[0]), '\\n')\n",
        "\n",
        "x = np.array([[[1 if dd > 0 else 0 for dd in m] for m in x] for x in train_x]) # change value to 1 if member train_x value more than 0\n",
        "y = to_categorical(train_y)\n",
        "\n",
        "print('After preprocess \\n', np.array(x[0]))\n",
        "print('After preprocess y \\n', np.array(y[0]))\n",
        "\n",
        "x = np.array([x.flatten() for x in x])"
      ],
      "metadata": {
        "id": "9EbFjMSH9iW4",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e99d0028-5c77-4f86-d074-b1c25abe4db1"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Before preprocess x \n",
            " [[  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   3  18  18  18 126 136 175  26 166 255 247 127   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0  30  36  94 154 170 253 253 253 253 253 225 172 253 242 195  64   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0  49 238 253 253 253 253 253 253 253 253 251  93  82  82  56  39   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0  18 219 253 253 253 253 253 198 182 247 241   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0  80 156 107 253 253 205  11   0  43 154   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0  14   1 154 253  90   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0 139 253 190   2   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0  11 190 253  70   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0  35 241 225 160 108   1   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0  81 240 253 253 119  25   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0  45 186 253 253 150  27   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0  16  93 252 253 187   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0 249 253 249  64   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0  46 130 183 253 253 207   2   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0  39 148 229 253 253 253 250 182   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0  24 114 221 253 253 253 253 201  78   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0  23  66 213 253 253 253 253 198  81   2   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0  18 171 219 253 253 253 253 195  80   9   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0  55 172 226 253 253 253 253 244 133  11   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0 136 253 253 253 212 135 132  16   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]] \n",
            "\n",
            "Before preprocess y \n",
            " 5 \n",
            "\n",
            "After preprocess \n",
            " [[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 0 1 1 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]]\n",
            "After preprocess y \n",
            " [0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# mlp\n",
        "input = x\n",
        "target = y\n",
        "\n",
        "# config\n",
        "num_input = input.shape[1]\n",
        "num_hidden = 30\n",
        "num_output = target.shape[1]\n",
        "\n",
        "# sigmoid func\n",
        "def sigmoid(x):\n",
        "  return 1/(1 + math.exp(-x))\n",
        "\n",
        "sig = np.vectorize(sigmoid)"
      ],
      "metadata": {
        "id": "1J7zkORUOih1"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# initialize weight and bias\n",
        "\n",
        "b_hidden = np.random.uniform(low=-1, high=1, size=(num_hidden))\n",
        "w_hidden = np.random.uniform(low=-1, high=1, size=(num_input, num_hidden))\n",
        "\n",
        "print('w_hidden:', w_hidden)\n",
        "print('b_hidden:', b_hidden)\n",
        "print('\\n')\n",
        "\n",
        "b_output = np.random.uniform(low=-1, high=1, size=(num_output))\n",
        "w_output = np.random.uniform(low=-1, high=1, size=(num_hidden, num_output))\n",
        "\n",
        "print('w_output:', w_output)\n",
        "print('b_output:', b_output)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7v7kBMMoPI0G",
        "outputId": "bc438ef6-0714-4d72-a09c-518145ba3983"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "w_hidden: [[-1.10105921  1.33876416  0.79701496 ... -1.42795032  1.00499773  2.63388975]\n",
            " [ 0.74323731 -2.36783271  2.17251955 ...  1.30343488  1.07295922 -2.65397517]\n",
            " [ 1.77627636 -2.24078257  1.08876232 ... -0.02107196 -2.64686288  2.21802082]\n",
            " ...\n",
            " [ 2.31037972 -0.63850262 -2.34327947 ... -2.17527563  2.59822912 -0.67784686]\n",
            " [ 1.12588806 -1.0142345   1.72269699 ...  2.68854723 -2.73673641 -1.83100721]\n",
            " [ 0.46389232  2.99216883  2.31534231 ... -1.51439684  2.4423046   1.79546742]]\n",
            "b_hidden: [-1.91146228  0.5801747  -2.08373757  1.5147185   0.0647356  -2.47075752 -2.43888343  1.63195488  2.01509913 -1.86217564  2.4990836   0.40947531 -1.2307795   1.78320794  0.75716391 -0.91086459\n",
            "  1.16047216 -0.20189308  1.9136347  -1.96399285  2.81402966  1.8289035  -1.57972789 -2.40574727  2.02464859 -0.6032486   2.01791313  2.60396313  0.37470634 -1.77008253]\n",
            "\n",
            "\n",
            "w_output: [[ 2.00922252 -0.88662764  1.92628029  0.93873662  1.92537676  1.80278828  1.91500643  1.43864596  1.14726396 -0.25708797]\n",
            " [-1.79468286 -2.65443475 -1.29925107 -1.29860271  2.37994775  2.26350244 -2.02990489 -2.0325803  -0.4225379  -2.25259726]\n",
            " [ 2.72035654  0.23573732  2.88087625 -0.6767165  -2.09339358  0.71033139  2.74656121  1.39955686  1.60733778 -0.86082938]\n",
            " [ 2.27604939 -2.87196252  2.6629479   0.99508652 -2.7654093  -2.35632191  0.69652995  0.70016241  0.3667181   0.99659971]\n",
            " [ 0.5142032  -2.64617137 -2.30822895 -1.53854699  1.53370816  1.03664953 -2.72415183 -2.68094255  0.21997769 -1.99779965]\n",
            " [ 2.79676125 -0.35007102  1.4201767  -0.05116258 -2.10591671  0.77576397  1.05144697 -0.9452445  -1.92781666  2.0581284 ]\n",
            " [-2.38434655 -2.88453683  0.83346391 -0.30731449 -1.92279237 -1.20658828 -2.64740784  1.3090494  -0.09903109  2.46261363]\n",
            " [ 1.81691877  2.23628816  0.59688676 -0.47883107 -0.21882176  2.91369928 -2.67475188  1.77422129 -2.66414614  0.906343  ]\n",
            " [-2.99221253 -1.73231141 -1.54802171  2.98341239  2.65016823  2.57537178  1.84614749  1.05108709 -0.24151434 -2.65181518]\n",
            " [-0.0274576  -1.98010566  2.0680059   0.83011725 -1.08770003  2.86903578  1.09643152 -1.17161531  1.99286034  0.32031369]\n",
            " [-1.94401151 -2.82471687  1.0434553   1.62349176  1.78058124  2.38486342 -2.75563799 -1.42603932  2.87437816  0.93978286]\n",
            " [-2.82466105 -2.83356588 -2.92797838 -0.60437402  2.62382076  2.74593817 -2.46182219 -0.77368855  2.15754576  1.08939904]\n",
            " [ 2.19015855  2.08527482 -1.64439673 -0.54020117 -1.0630927   2.26186235 -0.75982239 -1.31658009  2.83156623  1.64438825]\n",
            " [-2.98597276 -1.42409024 -2.19853395 -2.31459213 -2.69919944  1.67918987 -0.52841842  0.23067827  0.63774998  0.24615531]\n",
            " [-2.12900673  1.40489298 -1.91610246 -0.42039423 -1.25047682  1.20249836  1.64473679 -2.79095572  2.39333735 -1.79683646]\n",
            " [ 1.49451139  2.05888397  1.33570895 -1.58285228 -0.69320888 -2.60842138 -1.23669935  0.13299076  0.33653359  1.25276606]\n",
            " [ 2.91519443  2.70322535  1.50673963  0.69738095  1.84683091  1.9127235  -2.58593988 -2.04838343  1.72026612  2.63460466]\n",
            " [ 0.63842619 -2.91069245  2.55103541  0.68772414 -2.82446628 -0.77570061  1.82009674  2.98394133 -2.6930717   2.76332266]\n",
            " [ 1.21310687 -2.73735092 -0.17231029  1.81417288  0.61688629 -1.88648945 -2.59317456 -0.10168953  0.6700585   0.03173664]\n",
            " [-1.6172141   2.36107821 -2.42777507 -1.325956   -1.90077634  0.22339016  0.39312369 -1.26861857  2.06522902  0.42908992]\n",
            " [ 0.06520319  0.15498223 -0.53198776  0.00965094 -1.97412224 -2.57336203  0.24255457 -0.98321933 -1.74631415 -1.71704562]\n",
            " [ 0.2879831   0.36416495 -0.09667362  0.35628786 -1.20285304  0.8256977   2.03197756 -2.52956087 -2.2418145   1.45862087]\n",
            " [ 2.12042486  1.86431084 -0.59544833  0.44860413  2.74918311 -1.52797644 -2.74278888  2.76327516 -2.39269954 -2.34456678]\n",
            " [-2.51869325 -0.17705655 -2.31687906 -1.43714746  1.15373801  2.31875004 -1.96842091 -0.19024493 -1.56713598  1.4825626 ]\n",
            " [ 0.52135631 -2.89595345  2.41028503  0.60062994  2.02347656 -0.65605675 -2.99601269  1.92530037 -1.2825759   1.10770668]\n",
            " [-1.27364032  0.35087012 -1.9925852  -2.50648262 -2.43544547  1.6235022  -1.92243125 -0.3195787   1.36581686 -2.20363097]\n",
            " [-1.24536291  1.60758759  1.55157279 -0.03970029 -2.00793746 -0.51839277  2.20861184 -0.54566738  1.45278205  1.03708911]\n",
            " [ 1.32155965  2.11525498 -1.15809048 -0.31939726  2.08646307 -1.87539388 -0.08170709 -0.40425706 -0.90741039 -0.78979363]\n",
            " [-2.9554486   0.25754372  0.81429941 -0.81458368 -0.35378124  0.01171514 -1.2142531   2.64283607 -0.11022148  1.68522409]\n",
            " [ 2.13938799 -1.56856957  2.24102279  2.46553019  2.11054812  0.0378534   1.94588218 -2.379218   -0.05712844 -2.59850309]]\n",
            "b_output: [ 1.08955892 -1.64164896  0.97063098  1.15301223  1.96924129  1.41832547 -2.16540876  1.99018803 -1.56055379  1.18656481]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# loss and accuracy\n",
        "loss_values = []\n",
        "acc_values = []\n",
        "\n",
        "lr = 0.5\n",
        "epochs = 50\n",
        "\n",
        "for epoch in range(epochs):\n",
        "  MSE = 0\n",
        "  new_target = np.zeros(target.shape)\n",
        "\n",
        "  for idx, inp in enumerate(input): # feedforward\n",
        "    # hidden layer\n",
        "    o_hidden = np.matmul(input[idx], w_hidden) + b_hidden\n",
        "    o_hidden = sig(o_hidden)\n",
        "\n",
        "    # print('----------------')\n",
        "    # print(f'input[{idx}]:', input[idx], 'w_hidden:', w_hidden, 'b_hidden:', b_hidden)\n",
        "    # print('o_hidden:', o_hidden)\n",
        "    # print('\\n')\n",
        "\n",
        "    # output layer\n",
        "    o_output = np.matmul(o_hidden, w_output)\n",
        "    o_output = sig(o_output)\n",
        "\n",
        "    # print('o_hidden', o_hidden, 'w_output:', w_output, 'b_output:', b_output)\n",
        "    # print('o_output:', o_output)\n",
        "    # print('\\n')\n",
        "\n",
        "    # count error\n",
        "    error = target[idx] - o_output\n",
        "    MSE += ((np.sum(error**2)) / error.shape)\n",
        "\n",
        "    new_target[idx] = o_output.round()\n",
        "    eh = error @ w_output.T # error hidden\n",
        "\n",
        "    # print('error', error)\n",
        "    # print('eh', eh)\n",
        "    # print(f'new_target[{idx}]:', new_target[idx])\n",
        "    # print('\\n')\n",
        "\n",
        "\n",
        "    # backpropagation (using formula for sigmoid func)\n",
        "    w_output = w_output + (lr * (error * o_output * (1 - o_output)) * o_hidden[np.newaxis].T)\n",
        "    b_output = b_output + (lr * (error * o_output * (1 - o_output)))\n",
        "\n",
        "    w_hidden = w_hidden + (lr * (eh * o_hidden * (1 - o_hidden)) * input[idx][np.newaxis].T)\n",
        "    b_hidden = b_hidden + (lr * (eh * o_hidden * (1 - o_hidden)))\n",
        "\n",
        "  MSE = (MSE/input.shape[0])\n",
        "\n",
        "  delta = np.absolute(target - new_target)\n",
        "  acc = 1 - np.average([np.max(m) for m in delta])\n",
        "\n",
        "  loss_values.append(MSE)\n",
        "  acc_values.append(acc)\n",
        "\n",
        "  print('epoch:', epoch, '|', 'error:', MSE, 'acc:', acc)\n",
        "  # print('----------------')\n",
        "  # print('\\n')\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rv9ZZliMPaQG",
        "outputId": "52d5c713-899b-4f8a-badf-a46b1ab7295f"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "epoch: 0 | error: [0.09977917] acc: 0.08199999999999996\n",
            "epoch: 1 | error: [0.07375215] acc: 0.19999999999999996\n",
            "epoch: 2 | error: [0.0641719] acc: 0.32199999999999995\n",
            "epoch: 3 | error: [0.05873058] acc: 0.376\n",
            "epoch: 4 | error: [0.05282338] acc: 0.44599999999999995\n",
            "epoch: 5 | error: [0.05054596] acc: 0.482\n",
            "epoch: 6 | error: [0.04481915] acc: 0.54\n",
            "epoch: 7 | error: [0.03688199] acc: 0.6599999999999999\n",
            "epoch: 8 | error: [0.03949209] acc: 0.61\n",
            "epoch: 9 | error: [0.03533457] acc: 0.6659999999999999\n",
            "epoch: 10 | error: [0.03123976] acc: 0.72\n",
            "epoch: 11 | error: [0.02758384] acc: 0.752\n",
            "epoch: 12 | error: [0.02603103] acc: 0.772\n",
            "epoch: 13 | error: [0.0230709] acc: 0.802\n",
            "epoch: 14 | error: [0.02207826] acc: 0.81\n",
            "epoch: 15 | error: [0.02057807] acc: 0.8140000000000001\n",
            "epoch: 16 | error: [0.01724811] acc: 0.85\n",
            "epoch: 17 | error: [0.0154819] acc: 0.868\n",
            "epoch: 18 | error: [0.0139235] acc: 0.882\n",
            "epoch: 19 | error: [0.01351339] acc: 0.888\n",
            "epoch: 20 | error: [0.01413305] acc: 0.874\n",
            "epoch: 21 | error: [0.01407751] acc: 0.878\n",
            "epoch: 22 | error: [0.01499406] acc: 0.862\n",
            "epoch: 23 | error: [0.01209556] acc: 0.902\n",
            "epoch: 24 | error: [0.01076965] acc: 0.912\n",
            "epoch: 25 | error: [0.01189992] acc: 0.908\n",
            "epoch: 26 | error: [0.01162574] acc: 0.9\n",
            "epoch: 27 | error: [0.00998166] acc: 0.916\n",
            "epoch: 28 | error: [0.01381755] acc: 0.874\n",
            "epoch: 29 | error: [0.01012094] acc: 0.918\n",
            "epoch: 30 | error: [0.00879988] acc: 0.9299999999999999\n",
            "epoch: 31 | error: [0.01036949] acc: 0.908\n",
            "epoch: 32 | error: [0.01194357] acc: 0.886\n",
            "epoch: 33 | error: [0.00789457] acc: 0.924\n",
            "epoch: 34 | error: [0.00748303] acc: 0.9319999999999999\n",
            "epoch: 35 | error: [0.0078732] acc: 0.922\n",
            "epoch: 36 | error: [0.0078301] acc: 0.926\n",
            "epoch: 37 | error: [0.00728254] acc: 0.9339999999999999\n",
            "epoch: 38 | error: [0.00693949] acc: 0.938\n",
            "epoch: 39 | error: [0.00645649] acc: 0.94\n",
            "epoch: 40 | error: [0.00641548] acc: 0.944\n",
            "epoch: 41 | error: [0.00682287] acc: 0.9339999999999999\n",
            "epoch: 42 | error: [0.00668446] acc: 0.94\n",
            "epoch: 43 | error: [0.00638963] acc: 0.942\n",
            "epoch: 44 | error: [0.00615551] acc: 0.946\n",
            "epoch: 45 | error: [0.00588049] acc: 0.946\n",
            "epoch: 46 | error: [0.00582154] acc: 0.948\n",
            "epoch: 47 | error: [0.00623568] acc: 0.94\n",
            "epoch: 48 | error: [0.00665853] acc: 0.942\n",
            "epoch: 49 | error: [0.00607317] acc: 0.946\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plot1 = plt.figure(1)\n",
        "plt.plot(loss_values)\n",
        "plt.show()\n",
        "\n",
        "plot2 = plt.figure(2)\n",
        "plt.plot(acc_values)\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 843
        },
        "id": "MjklFLjOQiSs",
        "outputId": "678edc3b-455f-442d-c28d-34d7c6ea12b1"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "new_target_test = np.zeros(target.shape)\n",
        "\n",
        "# feedforward\n",
        "for idx, inp in enumerate(input):\n",
        "  o_hidden = np.matmul(input[idx], w_hidden) + b_hidden\n",
        "  o_hidden = sig(o_hidden)\n",
        "\n",
        "  o_output = np.matmul(o_hidden, w_output)\n",
        "  o_output = sig(o_output)\n",
        "\n",
        "  new_target_test[idx] = o_output.round()"
      ],
      "metadata": {
        "id": "C7ZNGhH0RODy"
      },
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "idx_test = 400\n",
        "\n",
        "reshape_input = input.reshape(num_data, 28, 28)\n",
        "print('reshape_input[idx_test]', reshape_input[idx_test])\n",
        "\n",
        "test_target = new_target_test[idx_test]\n",
        "print('test_target', test_target)\n",
        "max_index_row = np.argmax(test_target)\n",
        "print('max_index_row', max_index_row)\n",
        "\n",
        "plt.imshow(reshape_input[idx_test], cmap=plt.get_cmap('gray'))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 968
        },
        "id": "Vk7RjjS0SKpb",
        "outputId": "37165f4a-497a-495f-cb1a-a447d2cf2763"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "reshape_input[idx_test] [[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 1 1 1 0 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 1 1 1 1 1 1 1 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]]\n",
            "test_target [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]\n",
            "max_index_row 2\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7cbfa97ab5b0>"
            ]
          },
          "metadata": {},
          "execution_count": 12
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}