SimpleMLP.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "4ee14dcb-60f3-4eb4-b279-278f056c0abd",
   "metadata": {},
   "source": [
    "<p>Here, we try to simplify part of the code by implementing a simple and multi-layered neural network to have mentality.</p>\n",
    "\n",
    "* Train\n",
    "* Test\n",
    "* Save Weights"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c4ab6d7f-1b57-4e1d-accd-c79ef840a391",
   "metadata": {},
   "source": [
    "##### In all stages, we will use the existing dataset called </br>```train.csv``` for train network </br> ```test.csv``` for test and validate \n",
    "##### custom MNIST dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b8906838-1367-4594-ba4c-c3d27e17af09",
   "metadata": {},
   "source": [
    "#### imports "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "1c669192-2547-484c-a42a-c2266009df53",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np # matrix cals\n",
    "import pandas as pd # data analys\n",
    "from matplotlib import pyplot as plt # data Visualing\n",
    "import json # for save json format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "c5fb21cd-2490-4e26-a820-26255fc2e283",
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = pd.read_csv('train.csv') # load Train DataSet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f1614054-6ded-4f8c-a0e7-6c4044fa4d96",
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = np.array(dataset)\n",
    "np.random.shuffle(dataset) # shuffle before splitting into dev and training sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "a5e35b58-9170-4376-a6d8-995a0143e95d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(42000, 785)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "60ae214f-7718-4dcd-b32a-323e4737d53a",
   "metadata": {},
   "source": [
    "#### split train and test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "1e24dc43-2b0f-4f7e-8257-56d3a41f4b34",
   "metadata": {},
   "outputs": [],
   "source": [
    "tests = dataset[:1000].T # transpose Matrix | 1000 image test (785 row and 0-999 col)\n",
    "testY = tests[0] # split labels\n",
    "testX = tests[1:] # choise image data\n",
    "testX = testX / 255. # normalize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "1210cfd8-9391-441c-8933-79b13fc6a559",
   "metadata": {},
   "outputs": [],
   "source": [
    "trains = dataset[1000:].T # transpose Matrix | learning 41000 image (1000 to 42000 col)\n",
    "trainY = trains[0] # choise label\n",
    "trainX = trains[1:] # choise image data\n",
    "trainX = trainX / 255.  # normalize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "14ef6663-e303-42ed-afb0-8da50aff9d09",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((784, 1000), (784, 41000))"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(testX.shape, trainX.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d6a7a557-09c1-4938-8a37-03e7cf20181a",
   "metadata": {},
   "source": [
    "### MLP"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "ffa008f4-7a4e-41fd-8d40-f4dfc1e9ce20",
   "metadata": {},
   "outputs": [],
   "source": [
    "def init_params():\n",
    "    # each row = node\n",
    "    W1 = np.random.rand(100, 784) - 0.5  # 100 row and 784 col\n",
    "    b1 = np.random.rand(100, 1) - 0.5  # 100 row and 1 col\n",
    "    W2 = np.random.rand(10, 100) - 0.5  # 10 row and 100 col\n",
    "    b2 = np.random.rand(10, 1) - 0.5  # 10 row and 1 col\n",
    "    return W1, b1, W2, b2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "72bdd77f-5527-4802-894a-3dd426a2da8c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def ReLU(Z):\n",
    "    return np.maximum(Z, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "05de9ab0-c940-4dc1-91f3-51a3c4b17b3f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def softmax(Z):\n",
    "    return  np.exp(Z) / sum(np.exp(Z))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a4c20af4-6284-4244-957c-21ae2a3621be",
   "metadata": {},
   "outputs": [],
   "source": [
    "def ReLU_deriv(Z):\n",
    "    return Z > 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "22f82f87-4b3c-4c08-a579-427861e92d81",
   "metadata": {},
   "outputs": [],
   "source": [
    "def one_hot(Y):\n",
    "    one_hot_Y = np.zeros((Y.size, Y.max() + 1))\n",
    "    one_hot_Y[np.arange(Y.size), Y] = 1\n",
    "    return one_hot_Y.T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "e21026db-69e9-40cd-9777-d2cc584a65de",
   "metadata": {},
   "outputs": [],
   "source": [
    "def backward_prop(Z1, A1, Z2, A2, W1, W2, X, Y):\n",
    "    one_hot_Y = one_hot(Y)\n",
    "    dZ2 = A2 - one_hot_Y\n",
    "    dW2 = 1 / dataset.shape[0] * dZ2.dot(A1.T)\n",
    "    db2 = 1 / dataset.shape[0] * np.sum(dZ2)\n",
    "    dZ1 = W2.T.dot(dZ2) * ReLU_deriv(Z1)\n",
    "    dW1 = 1 / dataset.shape[0] * dZ1.dot(X.T)\n",
    "    db1 = 1 / dataset.shape[0] * np.sum(dZ1)\n",
    "    return dW1, db1, dW2, db2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "f0118fbc-4eed-4bee-a898-5cb03a72e4f0",
   "metadata": {},
   "outputs": [],
   "source": [
    "def update_params(W1, b1, W2, b2, dW1, db1, dW2, db2, learningRate):\n",
    "    W1 = W1 - learningRate * dW1\n",
    "    b1 = b1 - learningRate * db1    \n",
    "    W2 = W2 - learningRate * dW2  \n",
    "    b2 = b2 - learningRate * db2    \n",
    "    return W1, b1, W2, b2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "ce51670f-759d-428b-a21d-0e34a9e44809",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_accuracy(predictions, Y):\n",
    "    return np.sum(predictions == Y) / Y.size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "b33a7f7d-1e10-40ec-bf01-b8384646e2eb",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_predictions(A2):\n",
    "    return np.argmax(A2, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "dacf7826-bf6f-472b-9a37-1a209728434c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def forward_prop(W1, b1, W2, b2, X):\n",
    "    Z1 = W1.dot(X) + b1\n",
    "    A1 = ReLU(Z1)\n",
    "    Z2 = W2.dot(A1) + b2\n",
    "    A2 = softmax(Z2)\n",
    "    return Z1, A1, Z2, A2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "89e0d1d2-fa4c-4673-aeac-6d03c252cf61",
   "metadata": {},
   "outputs": [],
   "source": [
    "def gradient_descent(X, Y, learningRate, epoch):\n",
    "    W1, b1, W2, b2 = init_params()\n",
    "    for i in range(epoch):\n",
    "        Z1, A1, Z2, A2 = forward_prop(W1, b1, W2, b2, X)\n",
    "        dW1, db1, dW2, db2 = backward_prop(Z1, A1, Z2, A2, W1, W2, X, Y)\n",
    "        W1, b1, W2, b2 = update_params(W1, b1, W2, b2, dW1, db1, dW2, db2, learningRate)\n",
    "        if i % 10 == 0:\n",
    "            print(\"epoch: \", i)\n",
    "            predictions = get_predictions(A2)\n",
    "            print(\"Acrcy: \",get_accuracy(predictions, Y))\n",
    "    return W1, b1, W2, b2"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "88ce7acf-4bc1-4228-847f-b48af043bd93",
   "metadata": {},
   "source": [
    "### Train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "38199cf6-b33a-4c6b-ac9d-f0c3776670b8",
   "metadata": {},
   "outputs": [],
   "source": [
    "epoch = 200"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "463c009e-4942-4030-a2c7-c4ba1a3621c3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:  0\n",
      "Acrcy:  0.08724390243902438\n",
      "epoch:  10\n",
      "Acrcy:  0.3848536585365854\n",
      "epoch:  20\n",
      "Acrcy:  0.5304390243902439\n",
      "epoch:  30\n",
      "Acrcy:  0.6049024390243902\n",
      "epoch:  40\n",
      "Acrcy:  0.6538048780487805\n",
      "epoch:  50\n",
      "Acrcy:  0.6873902439024391\n",
      "epoch:  60\n",
      "Acrcy:  0.7138292682926829\n",
      "epoch:  70\n",
      "Acrcy:  0.7348292682926829\n",
      "epoch:  80\n",
      "Acrcy:  0.7529512195121951\n",
      "epoch:  90\n",
      "Acrcy:  0.7675365853658537\n",
      "epoch:  100\n",
      "Acrcy:  0.7797804878048781\n",
      "epoch:  110\n",
      "Acrcy:  0.7897804878048781\n",
      "epoch:  120\n",
      "Acrcy:  0.7988536585365854\n",
      "epoch:  130\n",
      "Acrcy:  0.8063902439024391\n",
      "epoch:  140\n",
      "Acrcy:  0.8123902439024391\n",
      "epoch:  150\n",
      "Acrcy:  0.8187560975609756\n",
      "epoch:  160\n",
      "Acrcy:  0.8241951219512195\n",
      "epoch:  170\n",
      "Acrcy:  0.8292195121951219\n",
      "epoch:  180\n",
      "Acrcy:  0.8344390243902439\n",
      "epoch:  190\n",
      "Acrcy:  0.8384878048780487\n"
     ]
    }
   ],
   "source": [
    "W1, b1, W2, b2 = gradient_descent(trainX, trainY, 0.10, epoch)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bca79e82-ad4d-45a9-9c12-d23f0f0502c1",
   "metadata": {},
   "source": [
    "### predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "21e7e046-3c94-474e-b2a5-92dcc001ea1f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_predictions(X, W1, b1, W2, b2):\n",
    "    _, _, _, A2 = forward_prop(W1, b1, W2, b2, X)\n",
    "    predictions = get_predictions(A2)\n",
    "    return predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "8941532b-a1c7-42d7-8ccb-53e8d2595567",
   "metadata": {},
   "outputs": [],
   "source": [
    "def test_prediction(index, W1, b1, W2, b2):\n",
    "    current_image = trainX[:, index, None]\n",
    "    prediction = make_predictions(trainX[:, index, None], W1, b1, W2, b2)\n",
    "    label = trainY[index]\n",
    "    print(\"Prediction: \", prediction)\n",
    "    print(\"Label: \", label)\n",
    "    current_image = current_image.reshape((28, 28)) * 255\n",
    "    plt.gray()\n",
    "    plt.imshow(current_image, interpolation='nearest')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "df040373-a471-4901-bdc9-5fbcb3f23eda",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction:  [1]\n",
      "Label:  1\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_prediction(0, W1, b1, W2, b2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "06d5cda3-d3e9-4953-bff2-105eb2a11ab5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction:  [6]\n",
      "Label:  6\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_prediction(3, W1, b1, W2, b2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "a6480bcb-7016-43d9-aee2-3f1ad7daa8f0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction:  [8]\n",
      "Label:  8\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_prediction(2, W1, b1, W2, b2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "eea700cb-627d-427b-8bb9-5db7138f8174",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction:  [8]\n",
      "Label:  8\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_prediction(4, W1, b1, W2, b2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "7df46525-cc00-443e-b99b-fa2b786d9f17",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.823"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_accuracy(make_predictions(testX, W1, b1, W2, b2), testY)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e8084438-a83c-4366-be91-8b46ea1c01b4",
   "metadata": {},
   "source": [
    "### accuracy for each number : "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "5c293f25-9a08-405d-af1d-60e0eedc86f1",
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = pd.read_csv('train.csv') # load Train DataSet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "a1c1deb8-7973-447e-a7f4-79477668fcd5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy 0 :  0.9439252336448598\n",
      "Accuracy 1 :  0.9270833333333334\n",
      "Accuracy 2 :  0.7983870967741935\n",
      "Accuracy 3 :  0.8888888888888888\n",
      "Accuracy 4 :  0.7843137254901961\n",
      "Accuracy 5 :  0.7191011235955056\n",
      "Accuracy 6 :  0.8762886597938144\n",
      "Accuracy 7 :  0.8761904761904762\n",
      "Accuracy 8 :  0.7311827956989247\n",
      "Accuracy 9 :  0.7938144329896907\n"
     ]
    }
   ],
   "source": [
    "for i in range(10):\n",
    "    values = np.array(dataset[:1000].query(f'label == {i}').T)   # peak vals\n",
    "    valueY = values[0]\n",
    "    valueX = values[1:] # choise image data\n",
    "    valueX = valueX / 255.  # normalize\n",
    "    print(f\"Accuracy {i} : \",get_accuracy(make_predictions(valueX, W1, b1, W2, b2), valueY))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a0880633-d5fe-4107-a592-f5109600394f",
   "metadata": {},
   "source": [
    "### save Weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "4a76833e-101e-434a-84ba-2073133f9aa4",
   "metadata": {},
   "outputs": [],
   "source": [
    "res = {\n",
    "\t\"w1\": W1.tolist(),\n",
    "\t\"w2\": W2.tolist(),\n",
    "    \"b1\": b1.tolist(),\n",
    "\t\"b2\": b2.tolist()\n",
    "}\n",
    "json_object = json.dumps(res, indent=4)\n",
    "with open(\"wgts\\simpleMLPWghts.json\", \"w\") as outfile:\n",
    "\toutfile.write(json_object)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2ad1410f-fa39-4ab5-b198-dbe6dcee1292",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}