Siamese_Ntw.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ebEGgSy2bdDp"
   },
   "source": [
    "# Siamese Network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "qc0w8-YjbiT3"
   },
   "source": [
    "## Importing the libraries and dependencies and drive\n",
    "The following cell block imports all the libraries, functions, methods that are being used inn this jupyter notebook.<br>\n",
    "The Deep Learning/Machine Learning Libarary being used here is the high level keras API of tensorflow 2.X.<Br>\n",
    "And the plotting is done with matplotlib or seaborn plotting packages.\n",
    "\n",
    "Along with this all the constants that will be common throughout the file are defined and declared here."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "qGMI8UW1sJ1f",
    "outputId": "0fff54f2-40cf-4504-c510-bbe33c65b853"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mounted at /content/drive\n"
     ]
    }
   ],
   "source": [
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "SyFQdFHhF87g"
   },
   "outputs": [],
   "source": [
    "import random, math, os, pickle, random, librosa\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow_datasets as tfds\n",
    "from itertools import combinations\n",
    "%matplotlib inline\n",
    "\n",
    "EPOCHS = 200\n",
    "BATCH_SIZE = 128\n",
    "IMG_SIZE = (28,28)\n",
    "SEED = 100\n",
    "CLASSES = 10\n",
    "LEARNING_RATE = 0.001\n",
    "ACCURACY_THRESHOLD = 0.98"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "IyUseb13cbtk"
   },
   "source": [
    "## Siamese network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "kSZ8uA2rdK8-"
   },
   "source": [
    "Load Training and Testing Data.\n",
    "\n",
    "data is made up of various audio files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Vz2f_Hw1dJY7",
    "outputId": "22bfdbd5-aa74-43fb-c1ca-2388edeead34"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(500, 16180)\n",
      "(200, 22631)\n"
     ]
    }
   ],
   "source": [
    "with open('/content/drive/MyDrive/data/h4_trs.pkl', 'rb') as train_file:\n",
    "    train_data = pickle.load(train_file)\n",
    "print(train_data.shape)\n",
    "\n",
    "with open('/content/drive/MyDrive/data/h4_tes.pkl', 'rb') as test_file:\n",
    "    test_data = pickle.load(test_file)\n",
    "print(test_data.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "gBst5vA8eoT8"
   },
   "source": [
    "### Create Positive pairs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "egNui2a0dnP7",
    "outputId": "13062ca9-b255-49b9-e71c-140a383328bd"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "45"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def create_positive_pairs(one_speaker, L=45):\n",
    "    batch = []\n",
    "    pairs = list(combinations(range(len(one_speaker)), 2))\n",
    "    one_pairs = random.sample(pairs, L)\n",
    "    for a,b in one_pairs:\n",
    "        a1 = np.zeros((45, 513))\n",
    "        b1 = np.zeros((45, 513))\n",
    "        stft_1 = np.abs(librosa.stft(one_speaker[a], n_fft=1024, hop_length=512)).T\n",
    "        a1[:stft_1.shape[0]] = stft_1\n",
    "        stft_2 = np.abs(librosa.stft(one_speaker[b], n_fft=1024, hop_length=512)).T\n",
    "        b1[:stft_2.shape[0]] = stft_2\n",
    "        batch.append([a1, b1])\n",
    "    return batch\n",
    "\n",
    "pos_batch = create_positive_pairs(train_data[10:20])\n",
    "len(pos_batch)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "x1NbcNCEfjmM"
   },
   "source": [
    "### Create Negative pairs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "MyZBQ4Cqk6WM",
    "outputId": "1e9786a8-65a1-4baa-c4aa-2782b83f28f4"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "45"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def create_negative_pairs(speaker_num, data, L=45):\n",
    "    batch = []\n",
    "    \n",
    "    start = speaker_num*10\n",
    "    end = start + 10     \n",
    "    speaker = random.choices(data[start: end], k=L)\n",
    "\n",
    "    neg = np.concatenate((data[:start], data[end:]))\n",
    "    not_speaker = random.choices(neg, k=L)\n",
    "\n",
    "    for speaker, not_speaker in zip(speaker, not_speaker):\n",
    "        x = np.abs(librosa.stft(speaker, n_fft=1024, hop_length=512)).T\n",
    "        stft_pos = np.zeros((45, 513))\n",
    "        stft_pos[:x.shape[0]] = x\n",
    "        y = np.abs(librosa.stft(not_speaker, n_fft=1024, hop_length=512)).T\n",
    "        stft_neg = np.zeros((45, 513))\n",
    "        stft_neg[:y.shape[0]] = y\n",
    "        batch.append([stft_pos, stft_neg])\n",
    "\n",
    "    return batch\n",
    "\n",
    "neg_batch = create_negative_pairs(5, train_data)\n",
    "len(neg_batch)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "jtb9FTLVgJwG"
   },
   "source": [
    "### Create training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "dxl4IgEOk-3f",
    "outputId": "3ff7f77f-084f-4374-b805-a2e13fde48d4"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((4500, 2, 45, 513), (4500,))"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batched_data = []\n",
    "labels = []\n",
    "\n",
    "for i in range(50):\n",
    "    pos_batch = create_positive_pairs(train_data[i*10:i*10+10])\n",
    "    pos_labels = list(np.ones((len(pos_batch))))\n",
    "\n",
    "    neg_batch = create_negative_pairs(i, train_data)\n",
    "    neg_labels = list(np.zeros((len(neg_batch))))\n",
    "    \n",
    "    batched_data += pos_batch + neg_batch\n",
    "    labels += pos_labels + neg_labels\n",
    "\n",
    "df_train = np.stack(batched_data)\n",
    "labels = np.stack(labels)\n",
    "df_train.shape, labels.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9s7cN6XMgMG1"
   },
   "source": [
    "### Create Testing/Validation data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "BPV1cUprOpvx",
    "outputId": "2c37b607-33a4-48f0-c421-f1375ce32b8a"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1800, 2, 45, 513), (1800,))"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_data_batched = []\n",
    "test_labels = []\n",
    "\n",
    "for i in range(20):\n",
    "    pos_batch = create_positive_pairs(test_data[i*10:i*10+10])\n",
    "    pos_labels = list(np.ones((len(pos_batch))))\n",
    "\n",
    "    neg_batch = create_negative_pairs(i, test_data)\n",
    "    neg_labels = list(np.zeros((len(neg_batch))))\n",
    "    \n",
    "    test_data_batched += pos_batch + neg_batch\n",
    "    test_labels += pos_labels + neg_labels\n",
    "\n",
    "df_test = np.stack(test_data_batched)\n",
    "test_labels = np.stack(test_labels)\n",
    "df_test.shape, test_labels.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "IRajJeDDgF_s"
   },
   "source": [
    "### Model Architecture & Summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "d9EISRrARa1T",
    "outputId": "a1527475-aaf2-4a3c-c2e5-917bc8668627"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_13\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " input_34 (InputLayer)          [(None, 45, 513)]    0           []                               \n",
      "                                                                                                  \n",
      " input_35 (InputLayer)          [(None, 45, 513)]    0           []                               \n",
      "                                                                                                  \n",
      " sequential_13 (Sequential)     (None, 45, 64)       7527024     ['input_34[0][0]',               \n",
      "                                                                  'input_35[0][0]']               \n",
      "                                                                                                  \n",
      " tf.math.multiply_11 (TFOpLambd  (None, 45, 64)      0           ['sequential_13[0][0]',          \n",
      " a)                                                               'sequential_13[1][0]']          \n",
      "                                                                                                  \n",
      " tf.math.reduce_sum_11 (TFOpLam  (None, 64)          0           ['tf.math.multiply_11[0][0]']    \n",
      " bda)                                                                                             \n",
      "                                                                                                  \n",
      " dense_35 (Dense)               (None, 1)            65          ['tf.math.reduce_sum_11[0][0]']  \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 7,527,089\n",
      "Trainable params: 7,525,037\n",
      "Non-trainable params: 2,052\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "def Siamese_model(shape = (45, 513)):\n",
    "    inp1 = tf.keras.Input(shape=shape)\n",
    "    inp2 = tf.keras.Input(shape=shape)\n",
    "\n",
    "    model = tf.keras.Sequential([\n",
    "                                tf.keras.Input(shape=shape),\n",
    "                                tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(513, return_sequences=True)),\n",
    "                                # tf.keras.layers.LSTM(513, return_sequences=True),\n",
    "                                tf.keras.layers.BatchNormalization(),\n",
    "                                tf.keras.layers.Dropout(0.2),\n",
    "                                tf.keras.layers.LSTM(513, return_sequences=True),\n",
    "                                #  tf.keras.layers.BatchNormalization(),\n",
    "                                tf.keras.layers.Dropout(0.2),\n",
    "                                tf.keras.layers.Dense(256, activation = 'relu'),\n",
    "                                tf.keras.layers.Dense(64, activation = 'relu'),\n",
    "                                #  tf.keras.layers.Lambda(lambda x: tf.where(x > 0.5, 1, 0))\n",
    "    ])\n",
    "\n",
    "    featsA = model(inp1)\n",
    "    featsB = model(inp2)\n",
    "\n",
    "    prediction = tf.reduce_sum(tf.multiply(featsA, featsB), axis = 1)\n",
    "\n",
    "    # distance = tf.keras.layers.Concatenate()([featsA, featsB])\n",
    "    outputs = tf.keras.layers.Dense(1, activation=\"sigmoid\")(prediction)\n",
    "\n",
    "    return tf.keras.models.Model(inputs=[inp1, inp2], outputs=outputs)\n",
    "\n",
    "model = Siamese_model()\n",
    "model.compile(loss=\"binary_crossentropy\", optimizer=\"adam\",\n",
    "\tmetrics=[\"accuracy\"])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Kga4nnhzgC2W"
   },
   "source": [
    "### Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "P9DTbnspUFHt",
    "outputId": "123b11b9-bc42-4fb6-c1b0-c2ea0221816f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "71/71 [==============================] - 18s 134ms/step - loss: 1.1913 - accuracy: 0.5547 - val_loss: 0.6800 - val_accuracy: 0.5644\n",
      "Epoch 2/100\n",
      "71/71 [==============================] - 7s 102ms/step - loss: 0.6079 - accuracy: 0.6698 - val_loss: 0.7593 - val_accuracy: 0.6200\n",
      "Epoch 3/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.5211 - accuracy: 0.7367 - val_loss: 0.6133 - val_accuracy: 0.7117\n",
      "Epoch 4/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.3978 - accuracy: 0.8176 - val_loss: 0.5869 - val_accuracy: 0.7606\n",
      "Epoch 5/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.3371 - accuracy: 0.8513 - val_loss: 0.5391 - val_accuracy: 0.7306\n",
      "Epoch 6/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.2558 - accuracy: 0.8980 - val_loss: 0.6645 - val_accuracy: 0.7394\n",
      "Epoch 7/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.2151 - accuracy: 0.9164 - val_loss: 0.7546 - val_accuracy: 0.6561\n",
      "Epoch 8/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.1948 - accuracy: 0.9207 - val_loss: 0.7136 - val_accuracy: 0.7689\n",
      "Epoch 9/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.1562 - accuracy: 0.9413 - val_loss: 0.6650 - val_accuracy: 0.7456\n",
      "Epoch 10/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.1262 - accuracy: 0.9484 - val_loss: 0.8056 - val_accuracy: 0.7478\n",
      "Epoch 11/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.1244 - accuracy: 0.9536 - val_loss: 0.9317 - val_accuracy: 0.7033\n",
      "Epoch 12/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0996 - accuracy: 0.9649 - val_loss: 1.1098 - val_accuracy: 0.7250\n",
      "Epoch 13/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0793 - accuracy: 0.9713 - val_loss: 1.0433 - val_accuracy: 0.7000\n",
      "Epoch 14/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0833 - accuracy: 0.9711 - val_loss: 0.8558 - val_accuracy: 0.7356\n",
      "Epoch 15/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.1197 - accuracy: 0.9582 - val_loss: 0.9268 - val_accuracy: 0.7406\n",
      "Epoch 16/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0724 - accuracy: 0.9751 - val_loss: 1.0711 - val_accuracy: 0.7061\n",
      "Epoch 17/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0691 - accuracy: 0.9764 - val_loss: 1.0058 - val_accuracy: 0.7078\n",
      "Epoch 18/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0617 - accuracy: 0.9789 - val_loss: 1.4169 - val_accuracy: 0.6828\n",
      "Epoch 19/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0774 - accuracy: 0.9747 - val_loss: 0.8917 - val_accuracy: 0.7711\n",
      "Epoch 20/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0896 - accuracy: 0.9693 - val_loss: 1.0869 - val_accuracy: 0.7072\n",
      "Epoch 21/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0605 - accuracy: 0.9822 - val_loss: 0.7671 - val_accuracy: 0.7383\n",
      "Epoch 22/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0445 - accuracy: 0.9862 - val_loss: 1.1880 - val_accuracy: 0.6606\n",
      "Epoch 23/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0225 - accuracy: 0.9938 - val_loss: 1.4220 - val_accuracy: 0.6911\n",
      "Epoch 24/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0128 - accuracy: 0.9947 - val_loss: 1.2339 - val_accuracy: 0.7406\n",
      "Epoch 25/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0291 - accuracy: 0.9922 - val_loss: 1.0510 - val_accuracy: 0.7489\n",
      "Epoch 26/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0336 - accuracy: 0.9902 - val_loss: 1.0904 - val_accuracy: 0.7294\n",
      "Epoch 27/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0490 - accuracy: 0.9847 - val_loss: 0.9288 - val_accuracy: 0.7250\n",
      "Epoch 28/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0268 - accuracy: 0.9922 - val_loss: 1.1908 - val_accuracy: 0.7350\n",
      "Epoch 29/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0300 - accuracy: 0.9896 - val_loss: 1.3701 - val_accuracy: 0.7289\n",
      "Epoch 30/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0282 - accuracy: 0.9907 - val_loss: 2.2995 - val_accuracy: 0.7011\n",
      "Epoch 31/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 0.0658 - accuracy: 0.9782 - val_loss: 1.1251 - val_accuracy: 0.7100\n",
      "Epoch 32/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0621 - accuracy: 0.9809 - val_loss: 2.0016 - val_accuracy: 0.6317\n",
      "Epoch 33/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 0.0442 - accuracy: 0.9853 - val_loss: 0.9976 - val_accuracy: 0.7067\n",
      "Epoch 34/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 0.0256 - accuracy: 0.9916 - val_loss: 1.0892 - val_accuracy: 0.7278\n",
      "Epoch 35/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 0.0365 - accuracy: 0.9889 - val_loss: 0.8972 - val_accuracy: 0.6811\n",
      "Epoch 36/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 0.0181 - accuracy: 0.9944 - val_loss: 1.5527 - val_accuracy: 0.6911\n",
      "Epoch 37/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 0.0164 - accuracy: 0.9960 - val_loss: 1.3962 - val_accuracy: 0.6983\n",
      "Epoch 38/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 0.0117 - accuracy: 0.9964 - val_loss: 1.7488 - val_accuracy: 0.6850\n",
      "Epoch 39/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0067 - accuracy: 0.9980 - val_loss: 1.9016 - val_accuracy: 0.6761\n",
      "Epoch 40/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0052 - accuracy: 0.9980 - val_loss: 1.7076 - val_accuracy: 0.6800\n",
      "Epoch 41/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 0.0030 - accuracy: 0.9991 - val_loss: 1.6679 - val_accuracy: 0.7033\n",
      "Epoch 42/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 9.0420e-04 - accuracy: 0.9996 - val_loss: 2.3710 - val_accuracy: 0.6622\n",
      "Epoch 43/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 2.0227e-04 - accuracy: 1.0000 - val_loss: 2.1499 - val_accuracy: 0.6806\n",
      "Epoch 44/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 1.1625e-04 - accuracy: 1.0000 - val_loss: 2.1403 - val_accuracy: 0.6867\n",
      "Epoch 45/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 4.9358e-05 - accuracy: 1.0000 - val_loss: 2.1747 - val_accuracy: 0.6867\n",
      "Epoch 46/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 3.9212e-05 - accuracy: 1.0000 - val_loss: 2.2206 - val_accuracy: 0.6839\n",
      "Epoch 47/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 4.0930e-05 - accuracy: 1.0000 - val_loss: 2.2715 - val_accuracy: 0.6817\n",
      "Epoch 48/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 2.8802e-05 - accuracy: 1.0000 - val_loss: 2.3139 - val_accuracy: 0.6772\n",
      "Epoch 49/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.6525e-05 - accuracy: 1.0000 - val_loss: 2.3182 - val_accuracy: 0.6828\n",
      "Epoch 50/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.1971e-05 - accuracy: 1.0000 - val_loss: 2.3389 - val_accuracy: 0.6822\n",
      "Epoch 51/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.1627e-05 - accuracy: 1.0000 - val_loss: 2.3637 - val_accuracy: 0.6822\n",
      "Epoch 52/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.2980e-05 - accuracy: 1.0000 - val_loss: 2.3745 - val_accuracy: 0.6833\n",
      "Epoch 53/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.7227e-05 - accuracy: 1.0000 - val_loss: 2.3918 - val_accuracy: 0.6828\n",
      "Epoch 54/100\n",
      "71/71 [==============================] - 7s 102ms/step - loss: 1.4433e-05 - accuracy: 1.0000 - val_loss: 2.4158 - val_accuracy: 0.6817\n",
      "Epoch 55/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.2579e-05 - accuracy: 1.0000 - val_loss: 2.4297 - val_accuracy: 0.6817\n",
      "Epoch 56/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.4492e-05 - accuracy: 1.0000 - val_loss: 2.4405 - val_accuracy: 0.6811\n",
      "Epoch 57/100\n",
      "71/71 [==============================] - 7s 102ms/step - loss: 1.3247e-05 - accuracy: 1.0000 - val_loss: 2.4467 - val_accuracy: 0.6822\n",
      "Epoch 58/100\n",
      "71/71 [==============================] - 7s 102ms/step - loss: 1.0133e-05 - accuracy: 1.0000 - val_loss: 2.4557 - val_accuracy: 0.6822\n",
      "Epoch 59/100\n",
      "71/71 [==============================] - 7s 102ms/step - loss: 8.6105e-06 - accuracy: 1.0000 - val_loss: 2.4755 - val_accuracy: 0.6822\n",
      "Epoch 60/100\n",
      "71/71 [==============================] - 7s 102ms/step - loss: 9.4122e-06 - accuracy: 1.0000 - val_loss: 2.4940 - val_accuracy: 0.6794\n",
      "Epoch 61/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 7.0581e-06 - accuracy: 1.0000 - val_loss: 2.5041 - val_accuracy: 0.6783\n",
      "Epoch 62/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 9.4251e-06 - accuracy: 1.0000 - val_loss: 2.5084 - val_accuracy: 0.6800\n",
      "Epoch 63/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 6.7801e-06 - accuracy: 1.0000 - val_loss: 2.5168 - val_accuracy: 0.6789\n",
      "Epoch 64/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 6.4543e-06 - accuracy: 1.0000 - val_loss: 2.5295 - val_accuracy: 0.6794\n",
      "Epoch 65/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 5.7578e-06 - accuracy: 1.0000 - val_loss: 2.5475 - val_accuracy: 0.6789\n",
      "Epoch 66/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 7.1657e-06 - accuracy: 1.0000 - val_loss: 2.5563 - val_accuracy: 0.6794\n",
      "Epoch 67/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 5.1561e-06 - accuracy: 1.0000 - val_loss: 2.5742 - val_accuracy: 0.6789\n",
      "Epoch 68/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 4.3415e-06 - accuracy: 1.0000 - val_loss: 2.5853 - val_accuracy: 0.6783\n",
      "Epoch 69/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 4.2870e-06 - accuracy: 1.0000 - val_loss: 2.5961 - val_accuracy: 0.6778\n",
      "Epoch 70/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 5.4900e-06 - accuracy: 1.0000 - val_loss: 2.6063 - val_accuracy: 0.6789\n",
      "Epoch 71/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 4.9894e-06 - accuracy: 1.0000 - val_loss: 2.6233 - val_accuracy: 0.6800\n",
      "Epoch 72/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 5.0479e-06 - accuracy: 1.0000 - val_loss: 2.6364 - val_accuracy: 0.6772\n",
      "Epoch 73/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 4.7152e-06 - accuracy: 1.0000 - val_loss: 2.6528 - val_accuracy: 0.6772\n",
      "Epoch 74/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 4.7848e-06 - accuracy: 1.0000 - val_loss: 2.6648 - val_accuracy: 0.6783\n",
      "Epoch 75/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 4.6448e-06 - accuracy: 1.0000 - val_loss: 2.6765 - val_accuracy: 0.6783\n",
      "Epoch 76/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 3.2138e-06 - accuracy: 1.0000 - val_loss: 2.6833 - val_accuracy: 0.6783\n",
      "Epoch 77/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 3.8075e-06 - accuracy: 1.0000 - val_loss: 2.6987 - val_accuracy: 0.6778\n",
      "Epoch 78/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 3.1851e-06 - accuracy: 1.0000 - val_loss: 2.7005 - val_accuracy: 0.6789\n",
      "Epoch 79/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.9840e-06 - accuracy: 1.0000 - val_loss: 2.7007 - val_accuracy: 0.6789\n",
      "Epoch 80/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 2.7269e-06 - accuracy: 1.0000 - val_loss: 2.7086 - val_accuracy: 0.6783\n",
      "Epoch 81/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.6009e-06 - accuracy: 1.0000 - val_loss: 2.7246 - val_accuracy: 0.6778\n",
      "Epoch 82/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.1583e-06 - accuracy: 1.0000 - val_loss: 2.7328 - val_accuracy: 0.6783\n",
      "Epoch 83/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.9977e-06 - accuracy: 1.0000 - val_loss: 2.7367 - val_accuracy: 0.6778\n",
      "Epoch 84/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 3.0317e-06 - accuracy: 1.0000 - val_loss: 2.7428 - val_accuracy: 0.6783\n",
      "Epoch 85/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.5712e-06 - accuracy: 1.0000 - val_loss: 2.7685 - val_accuracy: 0.6750\n",
      "Epoch 86/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.3978e-06 - accuracy: 1.0000 - val_loss: 2.7641 - val_accuracy: 0.6778\n",
      "Epoch 87/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.1097e-06 - accuracy: 1.0000 - val_loss: 2.7691 - val_accuracy: 0.6778\n",
      "Epoch 88/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.9748e-06 - accuracy: 1.0000 - val_loss: 2.7813 - val_accuracy: 0.6772\n",
      "Epoch 89/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 2.5576e-06 - accuracy: 1.0000 - val_loss: 2.7854 - val_accuracy: 0.6772\n",
      "Epoch 90/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.8324e-06 - accuracy: 1.0000 - val_loss: 2.7830 - val_accuracy: 0.6772\n",
      "Epoch 91/100\n",
      "71/71 [==============================] - 7s 102ms/step - loss: 1.6829e-06 - accuracy: 1.0000 - val_loss: 2.7961 - val_accuracy: 0.6778\n",
      "Epoch 92/100\n",
      "71/71 [==============================] - 7s 102ms/step - loss: 1.5079e-06 - accuracy: 1.0000 - val_loss: 2.8062 - val_accuracy: 0.6761\n",
      "Epoch 93/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.5102e-06 - accuracy: 1.0000 - val_loss: 2.8205 - val_accuracy: 0.6778\n",
      "Epoch 94/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.4047e-06 - accuracy: 1.0000 - val_loss: 2.8225 - val_accuracy: 0.6761\n",
      "Epoch 95/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.2752e-06 - accuracy: 1.0000 - val_loss: 2.8428 - val_accuracy: 0.6767\n",
      "Epoch 96/100\n",
      "71/71 [==============================] - 7s 104ms/step - loss: 1.3945e-06 - accuracy: 1.0000 - val_loss: 2.8545 - val_accuracy: 0.6761\n",
      "Epoch 97/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.5072e-06 - accuracy: 1.0000 - val_loss: 2.8578 - val_accuracy: 0.6756\n",
      "Epoch 98/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.4714e-06 - accuracy: 1.0000 - val_loss: 2.8706 - val_accuracy: 0.6761\n",
      "Epoch 99/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.2564e-06 - accuracy: 1.0000 - val_loss: 2.8855 - val_accuracy: 0.6756\n",
      "Epoch 100/100\n",
      "71/71 [==============================] - 7s 103ms/step - loss: 1.3281e-06 - accuracy: 1.0000 - val_loss: 2.9171 - val_accuracy: 0.6756\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f6ec9fcf290>"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit([df_train[:, 0], df_train[:, 1]], labels, \n",
    "          validation_data=([df_test[:,0], df_test[:,1]], test_labels), \n",
    "          batch_size=64, epochs=100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9m4gLtb7gAMn"
   },
   "source": [
    "### Plot the results\n",
    "Note:- Please ignore the X-label of the graphs. Its supposed to be \"Epochs\" but since the time is less - I labeled it D previously and am not able to run the whole model again due to shortage of time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 791
    },
    "id": "-Kcr0cYgT3Y6",
    "outputId": "521e5942-9ec8-44ff-aa4a-b7a0e479e8a6"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize = (10,6))\n",
    "plt.plot(model.history.history[\"accuracy\"])\n",
    "plt.plot(model.history.history[\"val_accuracy\"])\n",
    "plt.title(\"Accuracy of the model\")\n",
    "plt.xlabel('D')\n",
    "plt.ylabel(\"Accuracy\")\n",
    "plt.show()\n",
    "\n",
    "plt.figure(figsize = (10,6))\n",
    "plt.plot(model.history.history[\"loss\"])\n",
    "plt.plot(model.history.history[\"val_loss\"])\n",
    "plt.title(\"Loss of the model\")\n",
    "plt.xlabel('D')\n",
    "plt.ylabel(\"Loss\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_1VZaXq0fw7o"
   },
   "source": [
    "### Conclusion\n",
    "As it can be seen form the above that its quite evident that the model is over fitting to the training data. So the optimal choice of the epoch would be 20-30."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "09ZKp7GvXHU0",
    "outputId": "50985be9-5d4c-49d3-ef25-1cac2996d8a3"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "57/57 [==============================] - 1s 25ms/step - loss: 2.9171 - accuracy: 0.6756\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[2.917083501815796, 0.6755555272102356]"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate([df_test[:, 0], df_test[:, 1]] ,test_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "klDaxJC4XsTy",
    "outputId": "c6f5986a-aff6-4622-bb8d-30b11e4a9fdb"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1800, 45, 1)"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds = model.predict([df_test[:, 0], df_test[:, 1]])\n",
    "preds.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "WDFGnu2dYA5Z",
    "outputId": "50e2f096-1d75-464e-b8ad-65a84768b30e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1800, 2, 45, 513)"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test.shape"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "machine_shape": "hm",
   "name": "Assignment_4_DLS.ipynb",
   "provenance": []
  },
  "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}