2-14 딥러닝2 복습.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 비타민 14주 차 복습 문제\n",
    "## 딥러닝 2주 차 - 5기 4조\n",
    "\n",
    "### MNIST 손 글씨 분류하기\n",
    "#### *데이터 불러오기*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "id": "AF02lOVFwyvB",
    "outputId": "e026616c-465d-46b4-e9a7-3958aee1badf"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28)\n",
      "(10000, 28, 28)\n",
      "(60000,)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.datasets import mnist\n",
    "(x_train,y_train),(x_test,y_test)=mnist.load_data()\n",
    "#불러올 때 부터 train/test 데이터 셋과 feature/label을 따로 지정\n",
    "print(x_train.shape)\n",
    "print(x_test.shape)\n",
    "print(y_train.shape)\n",
    "#3D-Tensor의 이미지 데이터"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *label 값 분포 확인*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "id": "Q02GM9SVwyvo",
    "outputId": "9f32b21d-b5f5-4660-bb1b-2e6fb1238617"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train labels: {0: 5923, 1: 6742, 2: 5958, 3: 6131, 4: 5842, 5: 5421, 6: 5918, 7: 6265, 8: 5851, 9: 5949}\n",
      "Test labels: {0: 980, 1: 1135, 2: 1032, 3: 1010, 4: 982, 5: 892, 6: 958, 7: 1028, 8: 974, 9: 1009}\n"
     ]
    }
   ],
   "source": [
    "unique,counts=np.unique(y_train,return_counts=True)\n",
    "print('Train labels:',dict(zip(unique,counts)))\n",
    "unique,counts=np.unique(y_test,return_counts=True)\n",
    "print('Test labels:',dict(zip(unique,counts)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *이미지 표본 시각화*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "id": "G4NoEfJMwyvr",
    "outputId": "d0905d92-979e-4a78-832d-ef8f643bbb8b"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 144x144 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "# 4개의 임의 표본 추출하기\n",
    "indices=np.random.randint(0,x_train.shape[0],size=4)\n",
    "images=x_train[indices]\n",
    "labels=y_train[indices]\n",
    "# 4개 표본 그림그리기\n",
    "plt.figure(figsize=(2,2))\n",
    "for i in range(len(indices)):\n",
    "    plt.subplot(2,2,i+1)\n",
    "    image=images[i]\n",
    "    plt.imshow(image, cmap='gray')\n",
    "    plt.grid('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *피처 정규화 / 라벨 One-Hot 인코딩*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "id": "2z5qlhu3wyvu",
    "outputId": "d7f7a674-1bad-4084-b963-30b6b0af35db"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 10)\n"
     ]
    }
   ],
   "source": [
    "# one-hot coding을 위한 라이브러리 불러오기\n",
    "from keras.utils import np_utils\n",
    "num_features=784  #x_train.shape[1]*x_train.shape[2] a.k.a Flatten\n",
    "x_train=x_train.reshape(-1,num_features).astype('float32')\n",
    "x_test=x_test.reshape(-1,num_features).astype('float32')\n",
    "\n",
    "x_train=x_train/255.\n",
    "x_test=x_test/255.\n",
    "\n",
    "# 10개의 범주형 자료를 one-hot 벡터로 바꿔주기\n",
    "y_train=np_utils.to_categorical(y_train)\n",
    "y_test=np_utils.to_categorical(y_test)\n",
    "num_classes=y_train.shape[1]\n",
    "print(y_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 문제 1 : 조건에 맞게 코드를 작성해주세요\n",
    "#### *순차형 API 모델링*\n",
    "- Dense -> Relu -> Dropout -> Dense -> Relu -> Dense -> Softmax\n",
    "- 채널 숫자는 아래의 표 참고"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "Dg-uIcZWwyvy",
    "outputId": "d585c10f-02f1-40e2-f878-e9c8c092d747"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_9 (Dense)              (None, 256)               200960    \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 256)               65792     \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 10)                2570      \n",
      "=================================================================\n",
      "Total params: 269,322\n",
      "Trainable params: 269,322\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation, Dropout\n",
    "#2개의 은닉층과 1개의 출력층으로 구성된 MLP모형\n",
    "\n",
    "model=Sequential()\n",
    "####################################\n",
    "#######    fill this space   #######\n",
    "####################################\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *train / validation split*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "id": "-YFeEpDUwyv0",
    "outputId": "918b4227-4faa-4c14-d26a-dde25b8aa968"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(50000, 784)\n",
      "(50000, 10)\n"
     ]
    }
   ],
   "source": [
    "x_val=x_train[:10000]\n",
    "partial_x_train=x_train[10000:]\n",
    "y_val=y_train[:10000]\n",
    "partial_y_train=y_train[10000:]\n",
    "print(partial_x_train.shape)\n",
    "print(partial_y_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 문제 2 : 아래의 빈 칸을 채워주세요\n",
    "#### *모델 compile / fitting*\n",
    "- 손실 함수 : cross-entropy / 최적화 알고리즘 : adam / 평가 방법 : 정확도\n",
    "- 에폭 : 10 / 배치 사이즈 : 64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "id": "qlOvC6s8wyv4",
    "outputId": "3eae258b-a78b-4534-de4a-f1cbb8cac570"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "782/782 [==============================] - 10s 13ms/step - loss: 0.3387 - accuracy: 0.8976 - val_loss: 0.1521 - val_accuracy: 0.9548\n",
      "Epoch 2/10\n",
      "782/782 [==============================] - 9s 12ms/step - loss: 0.1631 - accuracy: 0.9497 - val_loss: 0.1119 - val_accuracy: 0.9675- loss:\n",
      "Epoch 3/10\n",
      "782/782 [==============================] - 8s 10ms/step - loss: 0.1300 - accuracy: 0.9590 - val_loss: 0.0956 - val_accuracy: 0.9699\n",
      "Epoch 4/10\n",
      "782/782 [==============================] - 5s 6ms/step - loss: 0.1114 - accuracy: 0.9650 - val_loss: 0.0851 - val_accuracy: 0.9743\n",
      "Epoch 5/10\n",
      "782/782 [==============================] - 6s 8ms/step - loss: 0.1021 - accuracy: 0.9683 - val_loss: 0.0770 - val_accuracy: 0.9775\n",
      "Epoch 6/10\n",
      "782/782 [==============================] - 7s 9ms/step - loss: 0.0883 - accuracy: 0.9712 - val_loss: 0.0788 - val_accuracy: 0.9772\n",
      "Epoch 7/10\n",
      "782/782 [==============================] - 7s 9ms/step - loss: 0.0854 - accuracy: 0.9724 - val_loss: 0.0754 - val_accuracy: 0.9786\n",
      "Epoch 8/10\n",
      "782/782 [==============================] - 7s 8ms/step - loss: 0.0762 - accuracy: 0.9761 - val_loss: 0.0835 - val_accuracy: 0.9770\n",
      "Epoch 9/10\n",
      "782/782 [==============================] - 8s 10ms/step - loss: 0.0707 - accuracy: 0.9768 - val_loss: 0.0738 - val_accuracy: 0.9784\n",
      "Epoch 10/10\n",
      "782/782 [==============================] - 6s 8ms/step - loss: 0.0682 - accuracy: 0.9771 - val_loss: 0.0718 - val_accuracy: 0.9796\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "########   fill this line   ######## \n",
    "history=model.fit(#### fill this blank ####)\n",
    "history_out=history.history\n",
    "history_out.keys()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *Loss 시각화*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "id": "NefNXHMZwyv6",
    "outputId": "6a938517-7c60-4080-e4b9-8f2c7bb8a328"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "loss=history_out['loss']\n",
    "loss_val=history_out['val_loss']\n",
    "accuracy=history_out['accuracy']\n",
    "accuracy_val=history_out['val_accuracy']\n",
    "plt.plot(loss,'bo',label='training loss')\n",
    "plt.plot( loss_val,'b', label='validation loss' )\n",
    "plt.title('Training and validation losss')\n",
    "plt.xlabel('epochs')\n",
    "plt.ylabel('loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *정확도 시각화*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "id": "0byp56_9wyv8",
    "outputId": "b502f52c-7865-4061-d074-44c3a14ca377"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.clf()\n",
    "plt.plot(accuracy,'bo',label='training loss')\n",
    "plt.plot(accuracy_val,'b', label='validation loss' )\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.xlabel('epochs')\n",
    "plt.ylabel('accuracy')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *test 셋 예측 / 평가*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "ifT9o3RwwywF",
    "outputId": "154f79a0-5364-4dc6-a7dd-60652ccae089"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 2s 26us/step\n",
      "10000/10000 [==============================] - 0s 25us/step\n",
      "[0.0016932068984431984, 0.9996166666666667]\n",
      "[0.07724655894196203, 0.9853]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6.194058e-20</td>\n",
       "      <td>1.543851e-16</td>\n",
       "      <td>5.002994e-12</td>\n",
       "      <td>5.879471e-12</td>\n",
       "      <td>1.921653e-17</td>\n",
       "      <td>2.172387e-20</td>\n",
       "      <td>8.632302e-35</td>\n",
       "      <td>1.000000e+00</td>\n",
       "      <td>1.505898e-15</td>\n",
       "      <td>1.402312e-10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1.773110e-15</td>\n",
       "      <td>1.343046e-16</td>\n",
       "      <td>1.000000e+00</td>\n",
       "      <td>2.149296e-15</td>\n",
       "      <td>3.389081e-21</td>\n",
       "      <td>1.079731e-18</td>\n",
       "      <td>7.891478e-17</td>\n",
       "      <td>2.729345e-17</td>\n",
       "      <td>1.753549e-15</td>\n",
       "      <td>1.040158e-20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1.003085e-12</td>\n",
       "      <td>9.999998e-01</td>\n",
       "      <td>1.764042e-07</td>\n",
       "      <td>2.180762e-11</td>\n",
       "      <td>4.792377e-10</td>\n",
       "      <td>6.744931e-13</td>\n",
       "      <td>8.021288e-09</td>\n",
       "      <td>8.970715e-08</td>\n",
       "      <td>1.060192e-08</td>\n",
       "      <td>1.556031e-15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1.000000e+00</td>\n",
       "      <td>2.679071e-22</td>\n",
       "      <td>8.096132e-12</td>\n",
       "      <td>6.897116e-14</td>\n",
       "      <td>3.153140e-16</td>\n",
       "      <td>2.451457e-12</td>\n",
       "      <td>1.956611e-10</td>\n",
       "      <td>2.660714e-17</td>\n",
       "      <td>2.974304e-18</td>\n",
       "      <td>1.921593e-11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1.328908e-09</td>\n",
       "      <td>3.842744e-12</td>\n",
       "      <td>2.796849e-11</td>\n",
       "      <td>8.706871e-13</td>\n",
       "      <td>9.999994e-01</td>\n",
       "      <td>4.950115e-14</td>\n",
       "      <td>9.592601e-10</td>\n",
       "      <td>4.728687e-07</td>\n",
       "      <td>6.603913e-11</td>\n",
       "      <td>1.569124e-07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              0             1             2             3             4  \\\n",
       "0  6.194058e-20  1.543851e-16  5.002994e-12  5.879471e-12  1.921653e-17   \n",
       "1  1.773110e-15  1.343046e-16  1.000000e+00  2.149296e-15  3.389081e-21   \n",
       "2  1.003085e-12  9.999998e-01  1.764042e-07  2.180762e-11  4.792377e-10   \n",
       "3  1.000000e+00  2.679071e-22  8.096132e-12  6.897116e-14  3.153140e-16   \n",
       "4  1.328908e-09  3.842744e-12  2.796849e-11  8.706871e-13  9.999994e-01   \n",
       "\n",
       "              5             6             7             8             9  \n",
       "0  2.172387e-20  8.632302e-35  1.000000e+00  1.505898e-15  1.402312e-10  \n",
       "1  1.079731e-18  7.891478e-17  2.729345e-17  1.753549e-15  1.040158e-20  \n",
       "2  6.744931e-13  8.021288e-09  8.970715e-08  1.060192e-08  1.556031e-15  \n",
       "3  2.451457e-12  1.956611e-10  2.660714e-17  2.974304e-18  1.921593e-11  \n",
       "4  4.950115e-14  9.592601e-10  4.728687e-07  6.603913e-11  1.569124e-07  "
      ]
     },
     "execution_count": 61,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "#loss,accuracy 평가\n",
    "result_train=model.evaluate(x_train,y_train)\n",
    "result_test=model.evaluate(x_test,y_test)\n",
    "print(result_train)\n",
    "print(result_test)\n",
    "\n",
    "#predict method\n",
    "pred=model.predict(x_test)\n",
    "pred=pd.DataFrame(pred)\n",
    "pred.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 문제 3 : 조건에 맞게 코드를 작성해주세요\n",
    "#### *함수형 API 모델링*\n",
    "- Layer 구성 및 채널 숫자는 위와 동일"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "id": "DoHCSLQCwywL"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"functional_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         [(None, 784)]             0         \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 256)               200960    \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_13 (Dense)             (None, 256)               65792     \n",
      "_________________________________________________________________\n",
      "dense_14 (Dense)             (None, 10)                2570      \n",
      "=================================================================\n",
      "Total params: 269,322\n",
      "Trainable params: 269,322\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.layers import Dense, Dropout, Input\n",
    "from keras.models import Model\n",
    "x=Input(shape=(num_features,))\n",
    "####################################\n",
    "#######    fill this space   #######\n",
    "####################################\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 문제 2와 동일 : 아래의 빈 칸을 채워주세요\n",
    "#### *모델 compile / fitting*\n",
    "- 손실 함수 : cross-entropy / 최적화 알고리즘 : adam / 평가 방법 : 정확도\n",
    "- 에폭 : 10 / 배치 사이즈 : 64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "938/938 [==============================] - 6s 7ms/step - loss: 0.3096 - acc: 0.9061\n",
      "Epoch 2/10\n",
      "938/938 [==============================] - 8s 9ms/step - loss: 0.1538 - acc: 0.9532\n",
      "Epoch 3/10\n",
      "938/938 [==============================] - 8s 8ms/step - loss: 0.1229 - acc: 0.9618\n",
      "Epoch 4/10\n",
      "938/938 [==============================] - 9s 9ms/step - loss: 0.1074 - acc: 0.9656A: 1s - los\n",
      "Epoch 5/10\n",
      "938/938 [==============================] - 7s 7ms/step - loss: 0.0969 - acc: 0.9692\n",
      "Epoch 6/10\n",
      "938/938 [==============================] - 7s 7ms/step - loss: 0.0843 - acc: 0.9730\n",
      "Epoch 7/10\n",
      "938/938 [==============================] - 7s 7ms/step - loss: 0.0780 - acc: 0.9743\n",
      "Epoch 8/10\n",
      "938/938 [==============================] - 6s 7ms/step - loss: 0.0726 - acc: 0.9769\n",
      "Epoch 9/10\n",
      "938/938 [==============================] - 7s 8ms/step - loss: 0.0664 - acc: 0.9789\n",
      "Epoch 10/10\n",
      "938/938 [==============================] - 6s 6ms/step - loss: 0.0635 - acc: 0.9790\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x1871aa7dd60>"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "########   fill this line   ######## \n",
    "model.fit(#### fill this blank ####)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 미니 배치 GD 클래스 구현하기\n",
    "#### *데이터 불러오기*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "shape of x : (442,)\n",
      "shape of y : (442,)\n"
     ]
    }
   ],
   "source": [
    "#PT의 예제 코드\n",
    "from sklearn.datasets import load_diabetes\n",
    "diabetes=load_diabetes()\n",
    "x=diabetes.data[:,2]\n",
    "y=diabetes.target\n",
    "print('shape of x :',x.shape)\n",
    "print('shape of y :',y.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 문제 4 : 아래의 빈 칸을 채워주세요\n",
    "#### *클래스 구현*\n",
    "- 미니 배치 사이즈 별 모수 최신화\n",
    "- loss : MSE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "class MiniBatch:\n",
    "    def __init__(self):\n",
    "        self.w=np.array([[1.0]])\n",
    "        self.b=np.array([[1.0]])\n",
    "        \n",
    "    def forpass(self,x):\n",
    "        ######## fill this space ########\n",
    "        return y_hat\n",
    "    \n",
    "    def backprop(self,x,err):\n",
    "        ######## fill this space ########\n",
    "        return w_grad, b_grad\n",
    "    \n",
    "    def fit(self,x,y,batch_size=32,lr=0.001,epochs=100):\n",
    "        x=x.reshape(-1,1)\n",
    "        y=y.reshape(-1,1)\n",
    "        for i in range(epochs):\n",
    "            indexes=np.random.permutation(np.arange(len(x)))\n",
    "            x=x[indexes]\n",
    "            y=y[indexes]\n",
    "            bins=len(x)//batch_size\n",
    "            if len(x)%batch_size:\n",
    "                bins+=1\n",
    "            for i in range(bins):\n",
    "                x_batch=######## fill this space ########\n",
    "                y_batch=######## fill this space ########\n",
    "                y_hat=######## fill this space ########\n",
    "                err=(y_batch-y_hat)/batch_size\n",
    "                w_grad,b_grad=######## fill this space ########\n",
    "                self.w+=lr*w_grad\n",
    "                self.b+=lr*b_grad"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *결과 시각화*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "neuron=MiniBatch()\n",
    "neuron.fit(x,y,epochs=10000)\n",
    "\n",
    "plt.scatter(x,y)\n",
    "pt1=(-0.1,-0.1*neuron.w[0]+neuron.b[0])\n",
    "pt2=(0.15,0.15*neuron.w[0]+neuron.b[0])\n",
    "plt.plot([pt1[0],pt2[0]],[pt1[1],pt2[1]],'r')\n",
    "plt.xlabel('x')\n",
    "plt.ylabel('y')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "name": "ch4_1.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}