Sudoku solver.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import de bibliothèques\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from tensorflow import keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100000 lignes et 2 colonnes\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>puzzle</th>\n",
       "      <th>solution</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0700000430400096108006349000940520003584600200...</td>\n",
       "      <td>6795182435437296188216349577943521863584617292...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>3010865040465210705000000014008000020803479000...</td>\n",
       "      <td>3719865248465213795924738614638197522853479167...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0483015603600080909106700030200009355090102006...</td>\n",
       "      <td>7483915623652487919126754834217869355894132766...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0083170000042051090000400703271609049014500000...</td>\n",
       "      <td>2983176457642851391539462783271689549814537266...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0408906300001368208007405190004670524500207002...</td>\n",
       "      <td>1428956379751368248367425193984671524513287962...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>99995</td>\n",
       "      <td>3257040900060037848470190021039028064891000036...</td>\n",
       "      <td>3257846919162537848476193521539728464891652736...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>99996</td>\n",
       "      <td>0906340706050094007840006939060028045180062302...</td>\n",
       "      <td>1926345786357894217845216939763528145184762392...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>99997</td>\n",
       "      <td>4050012000280070346300040980147609033001400858...</td>\n",
       "      <td>4953812671286975346372541985147689233691427858...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>99998</td>\n",
       "      <td>0502963000900502800423019750615408374301780007...</td>\n",
       "      <td>8572963141934572866423819752615498374351786927...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>99999</td>\n",
       "      <td>0005400300090020002540000000109240650608709004...</td>\n",
       "      <td>6715498328397124562543861797189243655638719244...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>100000 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  puzzle  \\\n",
       "0      0700000430400096108006349000940520003584600200...   \n",
       "1      3010865040465210705000000014008000020803479000...   \n",
       "2      0483015603600080909106700030200009355090102006...   \n",
       "3      0083170000042051090000400703271609049014500000...   \n",
       "4      0408906300001368208007405190004670524500207002...   \n",
       "...                                                  ...   \n",
       "99995  3257040900060037848470190021039028064891000036...   \n",
       "99996  0906340706050094007840006939060028045180062302...   \n",
       "99997  4050012000280070346300040980147609033001400858...   \n",
       "99998  0502963000900502800423019750615408374301780007...   \n",
       "99999  0005400300090020002540000000109240650608709004...   \n",
       "\n",
       "                                                solution  \n",
       "0      6795182435437296188216349577943521863584617292...  \n",
       "1      3719865248465213795924738614638197522853479167...  \n",
       "2      7483915623652487919126754834217869355894132766...  \n",
       "3      2983176457642851391539462783271689549814537266...  \n",
       "4      1428956379751368248367425193984671524513287962...  \n",
       "...                                                  ...  \n",
       "99995  3257846919162537848476193521539728464891652736...  \n",
       "99996  1926345786357894217845216939763528145184762392...  \n",
       "99997  4953812671286975346372541985147689233691427858...  \n",
       "99998  8572963141934572866423819752615498374351786927...  \n",
       "99999  6715498328397124562543861797189243655638719244...  \n",
       "\n",
       "[100000 rows x 2 columns]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Import des données (via Kaggle : https://www.kaggle.com/rohanrao/sudoku)\n",
    "\n",
    "nb_rows = 100000 # On charge 100 000 grilles\n",
    "data = pd.read_csv(\"sudoku.csv\", nrows=nb_rows)\n",
    "\n",
    "print(f'{data.shape[0]} lignes et {data.shape[1]} colonnes')\n",
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Mise en forme des grilles initiales\n",
    "\n",
    "for i in range(81):\n",
    "    data[i] = data['puzzle'].str[i].astype('int32')\n",
    "    \n",
    "X = data.drop(['puzzle', 'solution'], axis=1)\n",
    "X = np.array(X).reshape(X.shape[0], 9, 9, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Mise en forme des grilles remplies\n",
    "\n",
    "for i in range(81):\n",
    "    data[i] = data['solution'].str[i].astype('int32')\n",
    "    \n",
    "y = data.drop(['puzzle', 'solution'], axis=1)\n",
    "y = np.array(y).reshape(y.shape[0], 9, 9, 1)\n",
    "\n",
    "y = y-1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 80000 samples, validate on 20000 samples\n",
      "Epoch 1/20\n",
      "80000/80000 [==============================] - 405s 5ms/sample - loss: 1.0050 - val_loss: 0.5395\n",
      "Epoch 2/20\n",
      "80000/80000 [==============================] - 406s 5ms/sample - loss: 0.4803 - val_loss: 0.4671\n",
      "Epoch 3/20\n",
      "80000/80000 [==============================] - 401s 5ms/sample - loss: 0.4160 - val_loss: 0.4584\n",
      "Epoch 4/20\n",
      "80000/80000 [==============================] - 384s 5ms/sample - loss: 0.3503 - val_loss: 0.4871\n",
      "Epoch 5/20\n",
      "80000/80000 [==============================] - 384s 5ms/sample - loss: 0.2641 - val_loss: 0.5623\n"
     ]
    }
   ],
   "source": [
    "# Création du CNN\n",
    "\n",
    "model = keras.models.Sequential()\n",
    "\n",
    "model.add(keras.layers.Conv2D(64, kernel_size=(3,3), activation='relu', padding='same', input_shape=(9,9,1)))\n",
    "model.add(keras.layers.BatchNormalization())\n",
    "model.add(keras.layers.Conv2D(64, kernel_size=(3,3), activation='relu', padding='same'))\n",
    "model.add(keras.layers.BatchNormalization())\n",
    "model.add(keras.layers.Conv2D(128, kernel_size=(1,1), activation='relu', padding='same'))\n",
    "\n",
    "model.add(keras.layers.Flatten())\n",
    "model.add(keras.layers.Dense(81*9))\n",
    "model.add(keras.layers.Reshape((81, 9)))\n",
    "model.add(keras.layers.Activation('softmax')) \n",
    "\n",
    "model.compile(loss='sparse_categorical_crossentropy', optimizer='adam')\n",
    "\n",
    "early_stopping = keras.callbacks.EarlyStopping(monitor='val_loss',\n",
    "                                               patience=2, \n",
    "                                               restore_best_weights=True)\n",
    "\n",
    "history = model.fit(X, y, epochs=20, validation_split=0.2, callbacks=[early_stopping]) #0.3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred = model.predict_classes(X[0].reshape(1,9,9,1)).reshape(9,9)+1\n",
    "real = y[0].reshape(9,9)+1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ True,  True,  True, False, False,  True,  True,  True,  True],\n",
       "       [False,  True,  True,  True,  True,  True,  True,  True,  True],\n",
       "       [ True, False,  True,  True,  True,  True,  True,  True,  True],\n",
       "       [ True,  True,  True,  True,  True,  True,  True,  True, False],\n",
       "       [ True,  True,  True,  True,  True,  True, False,  True,  True],\n",
       "       [False, False, False,  True,  True,  True,  True,  True, False],\n",
       "       [ True,  True,  True,  True,  True,  True, False,  True,  True],\n",
       "       [ True, False,  True,  True,  True,  True,  True, False,  True],\n",
       "       [ True, False,  True, False,  True,  True,  True,  True,  True]])"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred == real"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparaison de grille"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[6, 7, 9, 2, 8, 8, 2, 4, 3],\n",
       "       [2, 4, 3, 7, 2, 9, 6, 1, 8],\n",
       "       [8, 1, 1, 6, 3, 4, 9, 5, 7],\n",
       "       [7, 9, 4, 3, 5, 2, 1, 8, 8],\n",
       "       [3, 5, 8, 4, 6, 1, 1, 2, 9],\n",
       "       [7, 2, 1, 8, 9, 7, 5, 3, 9],\n",
       "       [4, 8, 5, 2, 7, 6, 4, 9, 1],\n",
       "       [9, 3, 2, 1, 8, 3, 4, 6, 5],\n",
       "       [1, 1, 7, 3, 4, 5, 8, 6, 2]], dtype=int64)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred # Prédiction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[6, 7, 9, 5, 1, 8, 2, 4, 3],\n",
       "       [5, 4, 3, 7, 2, 9, 6, 1, 8],\n",
       "       [8, 2, 1, 6, 3, 4, 9, 5, 7],\n",
       "       [7, 9, 4, 3, 5, 2, 1, 8, 6],\n",
       "       [3, 5, 8, 4, 6, 1, 7, 2, 9],\n",
       "       [2, 1, 6, 8, 9, 7, 5, 3, 4],\n",
       "       [4, 8, 5, 2, 7, 6, 3, 9, 1],\n",
       "       [9, 6, 2, 1, 8, 3, 4, 7, 5],\n",
       "       [1, 3, 7, 9, 4, 5, 8, 6, 2]])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "real # Réalité"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8385337037037037% d'exactitude\n"
     ]
    }
   ],
   "source": [
    "acc = ((model.predict_classes(X).reshape(-1,9,9)+1 == y.reshape(-1,9,9)+1).sum())/(81*nb_rows)\n",
    "print(f'{acc}% d\\'exactitude') # 84 % d'exactitude"
   ]
  }
 ],
 "metadata": {
  "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": 2
}