Nadam.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Nadam1.ipynb",
      "provenance": [],
      "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/github/cs20m016/cs6910-Assignment-1/blob/main/Nadam.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SzgvqEwUVDnh"
      },
      "source": [
        "import numpy as np\r\n",
        "from matplotlib import pyplot as plt\r\n",
        "from matplotlib import gridspec as gridspec\r\n",
        "from random import randint\r\n",
        "from random import random\r\n",
        "from random import seed\r\n",
        "import math\r\n",
        "from sklearn.utils import shuffle"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t4AgIjoGaSpU"
      },
      "source": [
        "from keras.datasets import fashion_mnist"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bCvXdEeQaWlm",
        "outputId": "ab1e9140-2805-4c2a-d3bb-090d2079ad64"
      },
      "source": [
        "train_data,test_data=fashion_mnist.load_data()"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n",
            "32768/29515 [=================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz\n",
            "26427392/26421880 [==============================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n",
            "8192/5148 [===============================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz\n",
            "4423680/4422102 [==============================] - 0s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "27M0ILaNaXrD"
      },
      "source": [
        "train_data_mean = np.mean(train_data[0])\r\n",
        "train_data_stdev = np.std(train_data[0])\r\n",
        "train_data = ((train_data[0] - train_data_mean) / train_data_stdev, train_data[1])\r\n",
        "test_data = ((test_data[0] - train_data_mean) / train_data_stdev, test_data[1])\r\n",
        "train_X, trainY = train_data\r\n",
        "test_X, testY = test_data"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QAaCxeGAaaUd"
      },
      "source": [
        "train_X, trainY = shuffle(train_X, trainY)\r\n",
        "test_X, testY = shuffle(test_X, testY)"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mKxdOnLKac0W",
        "outputId": "0f0977d2-6ebd-42df-85dc-b1105de80b4e"
      },
      "source": [
        "print('Train: X=%s, y=%s' %(train_X.shape,trainY.shape))\r\n",
        "print('Test: X=%s, y=%s' %(test_X.shape,testY.shape))"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train: X=(60000, 28, 28), y=(60000,)\n",
            "Test: X=(10000, 28, 28), y=(10000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3VsTq-T6afCp"
      },
      "source": [
        "class_label=['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat', 'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xze8mLfJahvh"
      },
      "source": [
        "sample=[]\r\n",
        "for label in range(10):\r\n",
        "    for i in range(len(testY)):\r\n",
        "        if label==testY[i]:\r\n",
        "            sample.append(test_X[i])\r\n",
        "            break"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 118
        },
        "id": "vNPZLQZZajzt",
        "outputId": "3e67b7aa-9c22-46f2-b1d6-4688c811e342"
      },
      "source": [
        "fig=plt.figure(figsize=(15,15))\r\n",
        "rows=1\r\n",
        "columns=10\r\n",
        "for i in range(1, columns*rows +1):\r\n",
        "    tg=sample[i-1]\r\n",
        "    ax=fig.add_subplot(rows, columns, i)\r\n",
        "    plt.axis('off')\r\n",
        "    plt.imshow(tg,cmap=plt.cm.binary)\r\n",
        "    ax.title.set_text(class_label[i-1])\r\n",
        "plt.show()"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x1080 with 10 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-eXOr3geampu",
        "outputId": "07bcd717-4b7d-4974-fc1b-27b754bea948"
      },
      "source": [
        "trainX=[train_X[i].flatten() for i in range(len(train_X))]\r\n",
        "print(np.shape(trainX))\r\n",
        "print(np.shape(trainY))"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(60000, 784)\n",
            "(60000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sWuK5gb6ao9V",
        "outputId": "4e0a94f1-fcf2-40f0-f98e-e2c24ad73d91"
      },
      "source": [
        "testX=[test_X[i].flatten() for i in range(len(test_X))]\r\n",
        "print(np.shape(testX))\r\n",
        "print(np.shape(testY))"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(10000, 784)\n",
            "(10000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vLw3yQP0auK1",
        "outputId": "600b326e-5043-4765-b2c0-8eb13a4a6437"
      },
      "source": [
        "train_samples=len(trainX)\r\n",
        "test_samples=len(testX)\r\n",
        "xlen=len(trainX[0])\r\n",
        "print(xlen)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "784\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BxYjUlGsayKC"
      },
      "source": [
        "def initialize_weights(n_input, n_hidden_layer, n_output,neurons_hl):\r\n",
        "    W = list()\r\n",
        "    W.append([[np.random.normal() for j in range(n_input)]for i in range(neurons_hl)])\r\n",
        "    for i in range(n_hidden_layer-1):\r\n",
        "        W.append([[np.random.normal() for j in range(neurons_hl)]for i in range(neurons_hl)])\r\n",
        "    W.append([[np.random.normal() for j in range(neurons_hl)]for i in range(n_output)])\r\n",
        "    return W"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nWayAz-na4FN"
      },
      "source": [
        "def initialize_bias(n_input, n_hidden_layer, n_output,neurons_hl):\r\n",
        "    B=list()\r\n",
        "    for i in range(n_hidden_layer):\r\n",
        "        B.append([np.random.normal() for i in range(neurons_hl)])\r\n",
        "    B.append([np.random.normal() for i in range(n_output)])\r\n",
        "    return B"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dhOEEzqxa7YE"
      },
      "source": [
        "def sigmoid(b):\r\n",
        "  a=np.array(b,dtype=np.float128)\r\n",
        "  h=list()\r\n",
        "  for el in a:\r\n",
        "    h.append(1/(1+np.exp(-el)))\r\n",
        "  return h"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "he72AJ8da9xY"
      },
      "source": [
        "def softmax(b):\r\n",
        "  a=np.array(b,dtype=np.float128)\r\n",
        "  z = a - max(a)\r\n",
        "  h=list()\r\n",
        "  sum=0\r\n",
        "  for el in z:\r\n",
        "    sum+=np.exp(el)\r\n",
        "  for el in z:\r\n",
        "    h.append(np.exp(el)/sum)\r\n",
        "  return h"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v1uyHgAzawO8"
      },
      "source": [
        "L=3 #No of Layers\r\n",
        "N=32 #Size of each hidden layer\r\n",
        "K=10 #No of classes\r\n",
        "eta=0.0001 #Learning Rate"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5kK0RPujbA2x"
      },
      "source": [
        "def feed_forward(input,W,B,L):\r\n",
        "    H=list()\r\n",
        "    A=list()\r\n",
        "    H.append(input)\r\n",
        "    A.append([0])\r\n",
        "    for i in range(1,L):\r\n",
        "        a=B[i-1]+np.matmul(W[i-1],H[i-1])\r\n",
        "       # print(a)\r\n",
        "        A.append(a)\r\n",
        "       # print(h)\r\n",
        "        H.append(sigmoid(a))\r\n",
        "    a=B[L-1]+np.matmul(W[L-1],H[L-1])\r\n",
        "    A.append(a)\r\n",
        "   # print(a)\r\n",
        "    hL=softmax(a)\r\n",
        "   # print(hL)\r\n",
        "    H.append(hL)\r\n",
        "    return H,A,hL"
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LSxZy51nbDfy"
      },
      "source": [
        "def cmp(e):\r\n",
        "  g=1/(1+np.exp(-e))\r\n",
        "  return g*(1-g)"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NFda1VIRbGDd"
      },
      "source": [
        "def diff_sigmoid(a):\r\n",
        "    res=list()\r\n",
        "    for el in a:\r\n",
        "      res.append(cmp(el))\r\n",
        "    return res"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aBrk6bB6bIQw"
      },
      "source": [
        "def back_propogation(H,A,y_hat,label,W):\r\n",
        "    W_grad=list()\r\n",
        "    B_grad=list()\r\n",
        "    one_hot_y=np.zeros(K)\r\n",
        "    one_hot_y[label]+=1\r\n",
        "    ak_grad = y_hat-one_hot_y\r\n",
        "    for k in range(L,0,-1):\r\n",
        "        w_grad=np.matmul(np.matrix(ak_grad).T,np.matrix(H[k-1]))\r\n",
        "        W_grad.append(w_grad)\r\n",
        "        B_grad.append(ak_grad)\r\n",
        "        if k != 1:\r\n",
        "            h_grad=np.matmul(np.transpose(W[k-1]),ak_grad)\r\n",
        "            ak_grad=np.multiply(h_grad,diff_sigmoid(A[k-1]))\r\n",
        "    return W_grad,B_grad"
      ],
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fhegrOjOLdv4"
      },
      "source": [
        "def nadam():\r\n",
        "    t,e=0,0\r\n",
        "    epoch=5\r\n",
        "    batch_size=128\r\n",
        "    W= initialize_weights(xlen,L-1,K,N)\r\n",
        "    B= initialize_bias(xlen,L-1,K,N)\r\n",
        "    loss=list()\r\n",
        "    points=0\r\n",
        "    error=0.0\r\n",
        "    epsilon=1e-10\r\n",
        "    beta1=0.9\r\n",
        "    beta2=0.999\r\n",
        "    wgrad=list()\r\n",
        "    bgrad=list()\r\n",
        "    for l in range(L):\r\n",
        "        wgrad.append(np.zeros(shape=np.shape(W[l])).tolist())\r\n",
        "        bgrad.append(np.zeros(shape=np.shape(B[l])).tolist())\r\n",
        "    v_w=list()\r\n",
        "    v_b=list()\r\n",
        "    for l in range(L):\r\n",
        "        v_w.append(np.zeros(shape=np.shape(W[l])).tolist())\r\n",
        "        v_b.append(np.zeros(shape=np.shape(B[l])).tolist())\r\n",
        "    m_w=list()\r\n",
        "    m_b=list()\r\n",
        "    for l in range(L):\r\n",
        "        m_w.append(np.zeros(shape=np.shape(W[l])).tolist())\r\n",
        "        m_b.append(np.zeros(shape=np.shape(B[l])).tolist())\r\n",
        "    while (e<epoch):\r\n",
        "        for i in range(train_samples):\r\n",
        "            H,A,y_hat=feed_forward(trainX[i],W,B,L)\r\n",
        "            error+=(-math.log(y_hat[trainY[i]]))\r\n",
        "            W_grad,B_grad=back_propogation(H,A,y_hat,trainY[i],W)\r\n",
        "            W_grad=W_grad[::-1]\r\n",
        "            B_grad=B_grad[::-1]\r\n",
        "            for l in range(L):\r\n",
        "                wgrad[l]=(np.matrix(wgrad[l])+np.matrix(W_grad[l])).tolist()\r\n",
        "                bgrad[l]=(bgrad[l]+np.multiply(1,B_grad[l])).tolist()\r\n",
        "            points+=1\r\n",
        "            if(points%batch_size==0):\r\n",
        "                t+=1\r\n",
        "                points=0\r\n",
        "                loss.append(error/batch_size)\r\n",
        "                error=0.0\r\n",
        "                for l in range(L):\r\n",
        "                    m_w[l]=(np.multiply(beta1,m_w[l])+np.multiply((1-beta1),wgrad[l])).tolist()\r\n",
        "                    mwe=(np.divide(m_w[l],(1-(beta1**t)))).tolist()\r\n",
        "                    v_w[l]=(np.multiply(beta2,v_w[l])+np.multiply((1-beta2),np.square(wgrad[l]))).tolist()\r\n",
        "                    vwe=(np.divide(v_w[l],(1-(beta2**t)))).tolist()\r\n",
        "                    tmp=vwe.copy()\r\n",
        "                    new_t=np.add(epsilon,tmp).tolist()\r\n",
        "                    rl=np.reciprocal([[float(j) for j in i] for i in np.sqrt(new_t)]).tolist()\r\n",
        "                    mwe_up=(np.multiply(beta1,m_w[l])+np.multiply(((1-beta1)/(1-(beta1**t))),wgrad[l])).tolist()\r\n",
        "                    W[l]=(W[l]-np.multiply(mwe_up,np.multiply(eta,rl))).tolist()\r\n",
        "                    m_b[l]=(np.multiply(beta1,m_b[l])+np.multiply((1-beta1),bgrad[l])).tolist()\r\n",
        "                    mbe=(np.divide(m_b[l],(1-(beta1**t)))).tolist()\r\n",
        "                    v_b[l]=(np.multiply(beta2,v_b[l])+np.multiply((1-beta2),np.square(bgrad[l]))).tolist()\r\n",
        "                    vbe=(np.divide(v_b[l],(1-(beta2**t)))).tolist()\r\n",
        "                    tmp1=vbe.copy()\r\n",
        "                    new_t1=np.add(epsilon,tmp1).tolist()\r\n",
        "                    rl1=np.reciprocal([float(i) for i in np.sqrt(new_t1)]).tolist()\r\n",
        "                    mbe_up=(np.multiply(beta1,m_b[l])+np.multiply(((1-beta1)/(1-(beta1**t))),bgrad[l])).tolist()\r\n",
        "                    B[l]=np.subtract(B[l],np.multiply(mbe_up,np.multiply(eta,rl1))).tolist()              \r\n",
        "                wgrad=list()\r\n",
        "                bgrad=list()\r\n",
        "                for l in range(L):\r\n",
        "                    wgrad.append(np.zeros(shape=np.shape(W[l])).tolist())\r\n",
        "                    bgrad.append(np.zeros(shape=np.shape(B[l])).tolist()) \r\n",
        "        e=e+1\r\n",
        "        print('Epoch',e)\r\n",
        "    return W,B,loss"
      ],
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EaOBg5cZbSIG",
        "outputId": "76817002-ad1f-4c3e-d2f6-18b04d4ed1f2"
      },
      "source": [
        "W_r,B_r,loss=nadam()"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1\n",
            "Epoch 2\n",
            "Epoch 3\n",
            "Epoch 4\n",
            "Epoch 5\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 299
        },
        "id": "bugpAewFbS5Z",
        "outputId": "d363d13c-e22f-4057-efa2-e3d5465be01d"
      },
      "source": [
        "plt.plot(np.arange(1,len(loss)+1),loss)\r\n",
        "plt.title('Train loss- Nadam')"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5, 1.0, 'Train loss- Nadam')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 30
        },
        {
          "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"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J7bziNUBbZrB"
      },
      "source": [
        "#Prediction\r\n",
        "valid=0\r\n",
        "predict=[]\r\n",
        "for i in range(test_samples):\r\n",
        "    H,A,y_hat = feed_forward(testX[i],W_r,B_r,L)\r\n",
        "    class_predict = y_hat.index(max(y_hat))\r\n",
        "    predict.append(class_predict)\r\n",
        "    if(class_predict==testY[i]):\r\n",
        "      valid+=1"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "B_nfLx9vsSMc",
        "outputId": "9d2e180e-a37b-42bc-9559-dc4f6ece952b"
      },
      "source": [
        "valid"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "5855"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 32
        }
      ]
    }
  ]
}