├── .ipynb_checkpoints
    ├── CA_HOUSE_LINEAR_REGRESSION-checkpoint.ipynb
    ├── MNIST_LINEAR_CLASSIFIER-checkpoint.ipynb
    └── MNIST_MLN_CLASSIFIER-checkpoint.ipynb
├── CA_HOUSE_LINEAR_REGRESSION.ipynb
├── DIGIT_MLN_DROPOUT.ipynb
├── MNIST_AUTOENCODER.ipynb
├── MNIST_BILSTM.ipynb
├── MNIST_CNN.ipynb
├── MNIST_LINEAR_CLASSIFIER.ipynb
├── MNIST_LSTM.ipynb
├── MNIST_MLN_BN.ipynb
├── MNIST_MLN_CLASSIFIER.ipynb
├── MNIST_MLN_CLASSIFIER_DROPOUT.ipynb
├── README.md
└── datasets
    └── housing.csv


/.ipynb_checkpoints/CA_HOUSE_LINEAR_REGRESSION-checkpoint.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 1,
  6 |    "metadata": {
  7 |     "collapsed": true
  8 |    },
  9 |    "outputs": [],
 10 |    "source": [
 11 |     "import pandas as pd\n"
 12 |    ]
 13 |   },
 14 |   {
 15 |    "cell_type": "code",
 16 |    "execution_count": 2,
 17 |    "metadata": {
 18 |     "collapsed": true
 19 |    },
 20 |    "outputs": [],
 21 |    "source": [
 22 |     "housing_data = pd.read_csv('datasets/housing.csv')"
 23 |    ]
 24 |   },
 25 |   {
 26 |    "cell_type": "code",
 27 |    "execution_count": 3,
 28 |    "metadata": {},
 29 |    "outputs": [
 30 |     {
 31 |      "data": {
 32 |       "text/html": [
 33 |        "<div>\n",
 34 |        "<style>\n",
 35 |        "    .dataframe thead tr:only-child th {\n",
 36 |        "        text-align: right;\n",
 37 |        "    }\n",
 38 |        "\n",
 39 |        "    .dataframe thead th {\n",
 40 |        "        text-align: left;\n",
 41 |        "    }\n",
 42 |        "\n",
 43 |        "    .dataframe tbody tr th {\n",
 44 |        "        vertical-align: top;\n",
 45 |        "    }\n",
 46 |        "</style>\n",
 47 |        "<table border=\"1\" class=\"dataframe\">\n",
 48 |        "  <thead>\n",
 49 |        "    <tr style=\"text-align: right;\">\n",
 50 |        "      <th></th>\n",
 51 |        "      <th>longitude</th>\n",
 52 |        "      <th>latitude</th>\n",
 53 |        "      <th>housing_median_age</th>\n",
 54 |        "      <th>total_rooms</th>\n",
 55 |        "      <th>total_bedrooms</th>\n",
 56 |        "      <th>population</th>\n",
 57 |        "      <th>households</th>\n",
 58 |        "      <th>median_income</th>\n",
 59 |        "      <th>median_house_value</th>\n",
 60 |        "      <th>ocean_proximity</th>\n",
 61 |        "    </tr>\n",
 62 |        "  </thead>\n",
 63 |        "  <tbody>\n",
 64 |        "    <tr>\n",
 65 |        "      <th>0</th>\n",
 66 |        "      <td>-122.23</td>\n",
 67 |        "      <td>37.88</td>\n",
 68 |        "      <td>41.0</td>\n",
 69 |        "      <td>880.0</td>\n",
 70 |        "      <td>129.0</td>\n",
 71 |        "      <td>322.0</td>\n",
 72 |        "      <td>126.0</td>\n",
 73 |        "      <td>8.3252</td>\n",
 74 |        "      <td>452600.0</td>\n",
 75 |        "      <td>NEAR BAY</td>\n",
 76 |        "    </tr>\n",
 77 |        "    <tr>\n",
 78 |        "      <th>1</th>\n",
 79 |        "      <td>-122.22</td>\n",
 80 |        "      <td>37.86</td>\n",
 81 |        "      <td>21.0</td>\n",
 82 |        "      <td>7099.0</td>\n",
 83 |        "      <td>1106.0</td>\n",
 84 |        "      <td>2401.0</td>\n",
 85 |        "      <td>1138.0</td>\n",
 86 |        "      <td>8.3014</td>\n",
 87 |        "      <td>358500.0</td>\n",
 88 |        "      <td>NEAR BAY</td>\n",
 89 |        "    </tr>\n",
 90 |        "    <tr>\n",
 91 |        "      <th>2</th>\n",
 92 |        "      <td>-122.24</td>\n",
 93 |        "      <td>37.85</td>\n",
 94 |        "      <td>52.0</td>\n",
 95 |        "      <td>1467.0</td>\n",
 96 |        "      <td>190.0</td>\n",
 97 |        "      <td>496.0</td>\n",
 98 |        "      <td>177.0</td>\n",
 99 |        "      <td>7.2574</td>\n",
100 |        "      <td>352100.0</td>\n",
101 |        "      <td>NEAR BAY</td>\n",
102 |        "    </tr>\n",
103 |        "    <tr>\n",
104 |        "      <th>3</th>\n",
105 |        "      <td>-122.25</td>\n",
106 |        "      <td>37.85</td>\n",
107 |        "      <td>52.0</td>\n",
108 |        "      <td>1274.0</td>\n",
109 |        "      <td>235.0</td>\n",
110 |        "      <td>558.0</td>\n",
111 |        "      <td>219.0</td>\n",
112 |        "      <td>5.6431</td>\n",
113 |        "      <td>341300.0</td>\n",
114 |        "      <td>NEAR BAY</td>\n",
115 |        "    </tr>\n",
116 |        "    <tr>\n",
117 |        "      <th>4</th>\n",
118 |        "      <td>-122.25</td>\n",
119 |        "      <td>37.85</td>\n",
120 |        "      <td>52.0</td>\n",
121 |        "      <td>1627.0</td>\n",
122 |        "      <td>280.0</td>\n",
123 |        "      <td>565.0</td>\n",
124 |        "      <td>259.0</td>\n",
125 |        "      <td>3.8462</td>\n",
126 |        "      <td>342200.0</td>\n",
127 |        "      <td>NEAR BAY</td>\n",
128 |        "    </tr>\n",
129 |        "  </tbody>\n",
130 |        "</table>\n",
131 |        "</div>"
132 |       ],
133 |       "text/plain": [
134 |        "   longitude  latitude  housing_median_age  total_rooms  total_bedrooms  \\\n",
135 |        "0    -122.23     37.88                41.0        880.0           129.0   \n",
136 |        "1    -122.22     37.86                21.0       7099.0          1106.0   \n",
137 |        "2    -122.24     37.85                52.0       1467.0           190.0   \n",
138 |        "3    -122.25     37.85                52.0       1274.0           235.0   \n",
139 |        "4    -122.25     37.85                52.0       1627.0           280.0   \n",
140 |        "\n",
141 |        "   population  households  median_income  median_house_value ocean_proximity  \n",
142 |        "0       322.0       126.0         8.3252            452600.0        NEAR BAY  \n",
143 |        "1      2401.0      1138.0         8.3014            358500.0        NEAR BAY  \n",
144 |        "2       496.0       177.0         7.2574            352100.0        NEAR BAY  \n",
145 |        "3       558.0       219.0         5.6431            341300.0        NEAR BAY  \n",
146 |        "4       565.0       259.0         3.8462            342200.0        NEAR BAY  "
147 |       ]
148 |      },
149 |      "execution_count": 3,
150 |      "metadata": {},
151 |      "output_type": "execute_result"
152 |     }
153 |    ],
154 |    "source": [
155 |     "housing_data.head()\n",
156 |     "# 社区数据"
157 |    ]
158 |   },
159 |   {
160 |    "cell_type": "code",
161 |    "execution_count": 4,
162 |    "metadata": {},
163 |    "outputs": [
164 |     {
165 |      "name": "stdout",
166 |      "output_type": "stream",
167 |      "text": [
168 |       "<class 'pandas.core.frame.DataFrame'>\n",
169 |       "RangeIndex: 20640 entries, 0 to 20639\n",
170 |       "Data columns (total 10 columns):\n",
171 |       "longitude             20640 non-null float64\n",
172 |       "latitude              20640 non-null float64\n",
173 |       "housing_median_age    20640 non-null float64\n",
174 |       "total_rooms           20640 non-null float64\n",
175 |       "total_bedrooms        20433 non-null float64\n",
176 |       "population            20640 non-null float64\n",
177 |       "households            20640 non-null float64\n",
178 |       "median_income         20640 non-null float64\n",
179 |       "median_house_value    20640 non-null float64\n",
180 |       "ocean_proximity       20640 non-null object\n",
181 |       "dtypes: float64(9), object(1)\n",
182 |       "memory usage: 1.6+ MB\n"
183 |      ]
184 |     }
185 |    ],
186 |    "source": [
187 |     "housing_data.info()"
188 |    ]
189 |   },
190 |   {
191 |    "cell_type": "code",
192 |    "execution_count": 5,
193 |    "metadata": {},
194 |    "outputs": [
195 |     {
196 |      "name": "stdout",
197 |      "output_type": "stream",
198 |      "text": [
199 |       "<class 'pandas.core.frame.DataFrame'>\n",
200 |       "RangeIndex: 20640 entries, 0 to 20639\n",
201 |       "Data columns (total 10 columns):\n",
202 |       "longitude             20640 non-null float64\n",
203 |       "latitude              20640 non-null float64\n",
204 |       "housing_median_age    20640 non-null float64\n",
205 |       "total_rooms           20640 non-null float64\n",
206 |       "total_bedrooms        20640 non-null float64\n",
207 |       "population            20640 non-null float64\n",
208 |       "households            20640 non-null float64\n",
209 |       "median_income         20640 non-null float64\n",
210 |       "median_house_value    20640 non-null float64\n",
211 |       "ocean_proximity       20640 non-null object\n",
212 |       "dtypes: float64(9), object(1)\n",
213 |       "memory usage: 1.6+ MB\n"
214 |      ]
215 |     }
216 |    ],
217 |    "source": [
218 |     "median = housing_data['total_bedrooms'].median()\n",
219 |     "housing_data = housing_data.fillna(median)\n",
220 |     "\n",
221 |     "housing_data.info()"
222 |    ]
223 |   },
224 |   {
225 |    "cell_type": "code",
226 |    "execution_count": 6,
227 |    "metadata": {
228 |     "collapsed": true
229 |    },
230 |    "outputs": [],
231 |    "source": [
232 |     "housing_data = housing_data.drop('ocean_proximity', axis=1)\n",
233 |     "\n"
234 |    ]
235 |   },
236 |   {
237 |    "cell_type": "code",
238 |    "execution_count": 7,
239 |    "metadata": {},
240 |    "outputs": [
241 |     {
242 |      "name": "stdout",
243 |      "output_type": "stream",
244 |      "text": [
245 |       "<class 'pandas.core.frame.DataFrame'>\n",
246 |       "RangeIndex: 20640 entries, 0 to 20639\n",
247 |       "Data columns (total 9 columns):\n",
248 |       "longitude             20640 non-null float64\n",
249 |       "latitude              20640 non-null float64\n",
250 |       "housing_median_age    20640 non-null float64\n",
251 |       "total_rooms           20640 non-null float64\n",
252 |       "total_bedrooms        20640 non-null float64\n",
253 |       "population            20640 non-null float64\n",
254 |       "households            20640 non-null float64\n",
255 |       "median_income         20640 non-null float64\n",
256 |       "median_house_value    20640 non-null float64\n",
257 |       "dtypes: float64(9)\n",
258 |       "memory usage: 1.4 MB\n"
259 |      ]
260 |     }
261 |    ],
262 |    "source": [
263 |     "housing_data.info()"
264 |    ]
265 |   },
266 |   {
267 |    "cell_type": "code",
268 |    "execution_count": 8,
269 |    "metadata": {
270 |     "collapsed": true
271 |    },
272 |    "outputs": [],
273 |    "source": [
274 |     "\n",
275 |     "from sklearn.model_selection import train_test_split\n",
276 |     "\n",
277 |     "train_set, test_set = train_test_split(housing_data, test_size=0.2, random_state=33)"
278 |    ]
279 |   },
280 |   {
281 |    "cell_type": "code",
282 |    "execution_count": 9,
283 |    "metadata": {
284 |     "collapsed": true
285 |    },
286 |    "outputs": [],
287 |    "source": [
288 |     "y_train = train_set['median_house_value']\n",
289 |     "X_train = train_set.drop('median_house_value', axis=1)\n",
290 |     "\n",
291 |     "y_test = test_set['median_house_value']\n",
292 |     "X_test = test_set.drop('median_house_value', axis=1)\n",
293 |     "\n",
294 |     "from sklearn.linear_model import LinearRegression\n",
295 |     "\n",
296 |     "lr = LinearRegression()\n",
297 |     "\n",
298 |     "lr.fit(X_train, y_train)\n",
299 |     "\n",
300 |     "from sklearn.metrics.regression import mean_squared_error\n",
301 |     "\n",
302 |     "y_predict = lr.predict(X_test)"
303 |    ]
304 |   },
305 |   {
306 |    "cell_type": "code",
307 |    "execution_count": 10,
308 |    "metadata": {},
309 |    "outputs": [
310 |     {
311 |      "data": {
312 |       "text/plain": [
313 |        "69990.825384776152"
314 |       ]
315 |      },
316 |      "execution_count": 10,
317 |      "metadata": {},
318 |      "output_type": "execute_result"
319 |     }
320 |    ],
321 |    "source": [
322 |     "import numpy as np\n",
323 |     "print np.sqrt(mean_squared_error(y_test, y_predict))"
324 |    ]
325 |   },
326 |   {
327 |    "cell_type": "code",
328 |    "execution_count": 11,
329 |    "metadata": {
330 |     "collapsed": true
331 |    },
332 |    "outputs": [],
333 |    "source": [
334 |     "from sklearn.preprocessing import StandardScaler\n",
335 |     "\n",
336 |     "ss = StandardScaler()\n",
337 |     "\n",
338 |     "X_train = ss.fit_transform(X_train)\n",
339 |     "X_test = ss.fit_transform(X_test)"
340 |    ]
341 |   },
342 |   {
343 |    "cell_type": "code",
344 |    "execution_count": 13,
345 |    "metadata": {},
346 |    "outputs": [
347 |     {
348 |      "name": "stdout",
349 |      "output_type": "stream",
350 |      "text": [
351 |       "69620.6890544\n",
352 |       "69901.4152052\n"
353 |      ]
354 |     }
355 |    ],
356 |    "source": [
357 |     "lr = LinearRegression()\n",
358 |     "\n",
359 |     "lr.fit(X_train, y_train)\n",
360 |     "\n",
361 |     "from sklearn.metrics.regression import mean_squared_error\n",
362 |     "\n",
363 |     "\n",
364 |     "y_predict = lr.predict(X_train)\n",
365 |     "\n",
366 |     "import numpy as np\n",
367 |     "\n",
368 |     "\n",
369 |     "print np.sqrt(mean_squared_error(y_train, y_predict))\n",
370 |     "\n",
371 |     "y_predict = lr.predict(X_test)\n",
372 |     "\n",
373 |     "\n",
374 |     "print np.sqrt(mean_squared_error(y_test, y_predict))"
375 |    ]
376 |   },
377 |   {
378 |    "cell_type": "code",
379 |    "execution_count": 14,
380 |    "metadata": {},
381 |    "outputs": [
382 |     {
383 |      "name": "stdout",
384 |      "output_type": "stream",
385 |      "text": [
386 |       "0.0\n",
387 |       "79017.4707616\n"
388 |      ]
389 |     }
390 |    ],
391 |    "source": [
392 |     "from sklearn.tree import DecisionTreeRegressor\n",
393 |     "\n",
394 |     "dtr = DecisionTreeRegressor()\n",
395 |     "\n",
396 |     "dtr.fit(X_train, y_train)\n",
397 |     "\n",
398 |     "y_predict = dtr.predict(X_train)\n",
399 |     "\n",
400 |     "print np.sqrt(mean_squared_error(y_train, y_predict))\n",
401 |     "\n",
402 |     "from sklearn.metrics.regression import mean_squared_error\n",
403 |     "\n",
404 |     "y_predict = dtr.predict(X_test)\n",
405 |     "\n",
406 |     "import numpy as np\n",
407 |     "print np.sqrt(mean_squared_error(y_test, y_predict))"
408 |    ]
409 |   },
410 |   {
411 |    "cell_type": "code",
412 |    "execution_count": null,
413 |    "metadata": {
414 |     "collapsed": true
415 |    },
416 |    "outputs": [],
417 |    "source": []
418 |   }
419 |  ],
420 |  "metadata": {
421 |   "kernelspec": {
422 |    "display_name": "Python 2",
423 |    "language": "python",
424 |    "name": "python2"
425 |   },
426 |   "language_info": {
427 |    "codemirror_mode": {
428 |     "name": "ipython",
429 |     "version": 2
430 |    },
431 |    "file_extension": ".py",
432 |    "mimetype": "text/x-python",
433 |    "name": "python",
434 |    "nbconvert_exporter": "python",
435 |    "pygments_lexer": "ipython2",
436 |    "version": "2.7.13"
437 |   }
438 |  },
439 |  "nbformat": 4,
440 |  "nbformat_minor": 2
441 | }
442 | 


--------------------------------------------------------------------------------
/CA_HOUSE_LINEAR_REGRESSION.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 1,
  6 |    "metadata": {
  7 |     "collapsed": true
  8 |    },
  9 |    "outputs": [],
 10 |    "source": [
 11 |     "import pandas as pd\n"
 12 |    ]
 13 |   },
 14 |   {
 15 |    "cell_type": "code",
 16 |    "execution_count": 2,
 17 |    "metadata": {
 18 |     "collapsed": true
 19 |    },
 20 |    "outputs": [],
 21 |    "source": [
 22 |     "housing_data = pd.read_csv('datasets/housing.csv')"
 23 |    ]
 24 |   },
 25 |   {
 26 |    "cell_type": "code",
 27 |    "execution_count": 3,
 28 |    "metadata": {},
 29 |    "outputs": [
 30 |     {
 31 |      "data": {
 32 |       "text/html": [
 33 |        "<div>\n",
 34 |        "<style>\n",
 35 |        "    .dataframe thead tr:only-child th {\n",
 36 |        "        text-align: right;\n",
 37 |        "    }\n",
 38 |        "\n",
 39 |        "    .dataframe thead th {\n",
 40 |        "        text-align: left;\n",
 41 |        "    }\n",
 42 |        "\n",
 43 |        "    .dataframe tbody tr th {\n",
 44 |        "        vertical-align: top;\n",
 45 |        "    }\n",
 46 |        "</style>\n",
 47 |        "<table border=\"1\" class=\"dataframe\">\n",
 48 |        "  <thead>\n",
 49 |        "    <tr style=\"text-align: right;\">\n",
 50 |        "      <th></th>\n",
 51 |        "      <th>longitude</th>\n",
 52 |        "      <th>latitude</th>\n",
 53 |        "      <th>housing_median_age</th>\n",
 54 |        "      <th>total_rooms</th>\n",
 55 |        "      <th>total_bedrooms</th>\n",
 56 |        "      <th>population</th>\n",
 57 |        "      <th>households</th>\n",
 58 |        "      <th>median_income</th>\n",
 59 |        "      <th>median_house_value</th>\n",
 60 |        "      <th>ocean_proximity</th>\n",
 61 |        "    </tr>\n",
 62 |        "  </thead>\n",
 63 |        "  <tbody>\n",
 64 |        "    <tr>\n",
 65 |        "      <th>0</th>\n",
 66 |        "      <td>-122.23</td>\n",
 67 |        "      <td>37.88</td>\n",
 68 |        "      <td>41.0</td>\n",
 69 |        "      <td>880.0</td>\n",
 70 |        "      <td>129.0</td>\n",
 71 |        "      <td>322.0</td>\n",
 72 |        "      <td>126.0</td>\n",
 73 |        "      <td>8.3252</td>\n",
 74 |        "      <td>452600.0</td>\n",
 75 |        "      <td>NEAR BAY</td>\n",
 76 |        "    </tr>\n",
 77 |        "    <tr>\n",
 78 |        "      <th>1</th>\n",
 79 |        "      <td>-122.22</td>\n",
 80 |        "      <td>37.86</td>\n",
 81 |        "      <td>21.0</td>\n",
 82 |        "      <td>7099.0</td>\n",
 83 |        "      <td>1106.0</td>\n",
 84 |        "      <td>2401.0</td>\n",
 85 |        "      <td>1138.0</td>\n",
 86 |        "      <td>8.3014</td>\n",
 87 |        "      <td>358500.0</td>\n",
 88 |        "      <td>NEAR BAY</td>\n",
 89 |        "    </tr>\n",
 90 |        "    <tr>\n",
 91 |        "      <th>2</th>\n",
 92 |        "      <td>-122.24</td>\n",
 93 |        "      <td>37.85</td>\n",
 94 |        "      <td>52.0</td>\n",
 95 |        "      <td>1467.0</td>\n",
 96 |        "      <td>190.0</td>\n",
 97 |        "      <td>496.0</td>\n",
 98 |        "      <td>177.0</td>\n",
 99 |        "      <td>7.2574</td>\n",
100 |        "      <td>352100.0</td>\n",
101 |        "      <td>NEAR BAY</td>\n",
102 |        "    </tr>\n",
103 |        "    <tr>\n",
104 |        "      <th>3</th>\n",
105 |        "      <td>-122.25</td>\n",
106 |        "      <td>37.85</td>\n",
107 |        "      <td>52.0</td>\n",
108 |        "      <td>1274.0</td>\n",
109 |        "      <td>235.0</td>\n",
110 |        "      <td>558.0</td>\n",
111 |        "      <td>219.0</td>\n",
112 |        "      <td>5.6431</td>\n",
113 |        "      <td>341300.0</td>\n",
114 |        "      <td>NEAR BAY</td>\n",
115 |        "    </tr>\n",
116 |        "    <tr>\n",
117 |        "      <th>4</th>\n",
118 |        "      <td>-122.25</td>\n",
119 |        "      <td>37.85</td>\n",
120 |        "      <td>52.0</td>\n",
121 |        "      <td>1627.0</td>\n",
122 |        "      <td>280.0</td>\n",
123 |        "      <td>565.0</td>\n",
124 |        "      <td>259.0</td>\n",
125 |        "      <td>3.8462</td>\n",
126 |        "      <td>342200.0</td>\n",
127 |        "      <td>NEAR BAY</td>\n",
128 |        "    </tr>\n",
129 |        "  </tbody>\n",
130 |        "</table>\n",
131 |        "</div>"
132 |       ],
133 |       "text/plain": [
134 |        "   longitude  latitude  housing_median_age  total_rooms  total_bedrooms  \\\n",
135 |        "0    -122.23     37.88                41.0        880.0           129.0   \n",
136 |        "1    -122.22     37.86                21.0       7099.0          1106.0   \n",
137 |        "2    -122.24     37.85                52.0       1467.0           190.0   \n",
138 |        "3    -122.25     37.85                52.0       1274.0           235.0   \n",
139 |        "4    -122.25     37.85                52.0       1627.0           280.0   \n",
140 |        "\n",
141 |        "   population  households  median_income  median_house_value ocean_proximity  \n",
142 |        "0       322.0       126.0         8.3252            452600.0        NEAR BAY  \n",
143 |        "1      2401.0      1138.0         8.3014            358500.0        NEAR BAY  \n",
144 |        "2       496.0       177.0         7.2574            352100.0        NEAR BAY  \n",
145 |        "3       558.0       219.0         5.6431            341300.0        NEAR BAY  \n",
146 |        "4       565.0       259.0         3.8462            342200.0        NEAR BAY  "
147 |       ]
148 |      },
149 |      "execution_count": 3,
150 |      "metadata": {},
151 |      "output_type": "execute_result"
152 |     }
153 |    ],
154 |    "source": [
155 |     "housing_data.head()\n",
156 |     "# 社区数据"
157 |    ]
158 |   },
159 |   {
160 |    "cell_type": "code",
161 |    "execution_count": 4,
162 |    "metadata": {},
163 |    "outputs": [
164 |     {
165 |      "name": "stdout",
166 |      "output_type": "stream",
167 |      "text": [
168 |       "<class 'pandas.core.frame.DataFrame'>\n",
169 |       "RangeIndex: 20640 entries, 0 to 20639\n",
170 |       "Data columns (total 10 columns):\n",
171 |       "longitude             20640 non-null float64\n",
172 |       "latitude              20640 non-null float64\n",
173 |       "housing_median_age    20640 non-null float64\n",
174 |       "total_rooms           20640 non-null float64\n",
175 |       "total_bedrooms        20433 non-null float64\n",
176 |       "population            20640 non-null float64\n",
177 |       "households            20640 non-null float64\n",
178 |       "median_income         20640 non-null float64\n",
179 |       "median_house_value    20640 non-null float64\n",
180 |       "ocean_proximity       20640 non-null object\n",
181 |       "dtypes: float64(9), object(1)\n",
182 |       "memory usage: 1.6+ MB\n"
183 |      ]
184 |     }
185 |    ],
186 |    "source": [
187 |     "housing_data.info()"
188 |    ]
189 |   },
190 |   {
191 |    "cell_type": "code",
192 |    "execution_count": 5,
193 |    "metadata": {},
194 |    "outputs": [
195 |     {
196 |      "name": "stdout",
197 |      "output_type": "stream",
198 |      "text": [
199 |       "<class 'pandas.core.frame.DataFrame'>\n",
200 |       "RangeIndex: 20640 entries, 0 to 20639\n",
201 |       "Data columns (total 10 columns):\n",
202 |       "longitude             20640 non-null float64\n",
203 |       "latitude              20640 non-null float64\n",
204 |       "housing_median_age    20640 non-null float64\n",
205 |       "total_rooms           20640 non-null float64\n",
206 |       "total_bedrooms        20640 non-null float64\n",
207 |       "population            20640 non-null float64\n",
208 |       "households            20640 non-null float64\n",
209 |       "median_income         20640 non-null float64\n",
210 |       "median_house_value    20640 non-null float64\n",
211 |       "ocean_proximity       20640 non-null object\n",
212 |       "dtypes: float64(9), object(1)\n",
213 |       "memory usage: 1.6+ MB\n"
214 |      ]
215 |     }
216 |    ],
217 |    "source": [
218 |     "median = housing_data['total_bedrooms'].median()\n",
219 |     "housing_data = housing_data.fillna(median)\n",
220 |     "\n",
221 |     "housing_data.info()"
222 |    ]
223 |   },
224 |   {
225 |    "cell_type": "code",
226 |    "execution_count": 6,
227 |    "metadata": {
228 |     "collapsed": true
229 |    },
230 |    "outputs": [],
231 |    "source": [
232 |     "housing_data = housing_data.drop('ocean_proximity', axis=1)\n",
233 |     "\n"
234 |    ]
235 |   },
236 |   {
237 |    "cell_type": "code",
238 |    "execution_count": 7,
239 |    "metadata": {},
240 |    "outputs": [
241 |     {
242 |      "name": "stdout",
243 |      "output_type": "stream",
244 |      "text": [
245 |       "<class 'pandas.core.frame.DataFrame'>\n",
246 |       "RangeIndex: 20640 entries, 0 to 20639\n",
247 |       "Data columns (total 9 columns):\n",
248 |       "longitude             20640 non-null float64\n",
249 |       "latitude              20640 non-null float64\n",
250 |       "housing_median_age    20640 non-null float64\n",
251 |       "total_rooms           20640 non-null float64\n",
252 |       "total_bedrooms        20640 non-null float64\n",
253 |       "population            20640 non-null float64\n",
254 |       "households            20640 non-null float64\n",
255 |       "median_income         20640 non-null float64\n",
256 |       "median_house_value    20640 non-null float64\n",
257 |       "dtypes: float64(9)\n",
258 |       "memory usage: 1.4 MB\n"
259 |      ]
260 |     }
261 |    ],
262 |    "source": [
263 |     "housing_data.info()"
264 |    ]
265 |   },
266 |   {
267 |    "cell_type": "code",
268 |    "execution_count": 8,
269 |    "metadata": {
270 |     "collapsed": true
271 |    },
272 |    "outputs": [],
273 |    "source": [
274 |     "\n",
275 |     "from sklearn.model_selection import train_test_split\n",
276 |     "\n",
277 |     "train_set, test_set = train_test_split(housing_data, test_size=0.2, random_state=33)"
278 |    ]
279 |   },
280 |   {
281 |    "cell_type": "code",
282 |    "execution_count": 9,
283 |    "metadata": {
284 |     "collapsed": true
285 |    },
286 |    "outputs": [],
287 |    "source": [
288 |     "y_train = train_set['median_house_value']\n",
289 |     "X_train = train_set.drop('median_house_value', axis=1)\n",
290 |     "\n",
291 |     "y_test = test_set['median_house_value']\n",
292 |     "X_test = test_set.drop('median_house_value', axis=1)\n",
293 |     "\n",
294 |     "from sklearn.linear_model import LinearRegression\n",
295 |     "\n",
296 |     "lr = LinearRegression()\n",
297 |     "\n",
298 |     "lr.fit(X_train, y_train)\n",
299 |     "\n",
300 |     "from sklearn.metrics.regression import mean_squared_error\n",
301 |     "\n",
302 |     "y_predict = lr.predict(X_test)"
303 |    ]
304 |   },
305 |   {
306 |    "cell_type": "code",
307 |    "execution_count": 10,
308 |    "metadata": {},
309 |    "outputs": [
310 |     {
311 |      "data": {
312 |       "text/plain": [
313 |        "69990.825384776152"
314 |       ]
315 |      },
316 |      "execution_count": 10,
317 |      "metadata": {},
318 |      "output_type": "execute_result"
319 |     }
320 |    ],
321 |    "source": [
322 |     "import numpy as np\n",
323 |     "print np.sqrt(mean_squared_error(y_test, y_predict))"
324 |    ]
325 |   },
326 |   {
327 |    "cell_type": "code",
328 |    "execution_count": 11,
329 |    "metadata": {
330 |     "collapsed": true
331 |    },
332 |    "outputs": [],
333 |    "source": [
334 |     "from sklearn.preprocessing import StandardScaler\n",
335 |     "\n",
336 |     "ss = StandardScaler()\n",
337 |     "\n",
338 |     "X_train = ss.fit_transform(X_train)\n",
339 |     "X_test = ss.fit_transform(X_test)"
340 |    ]
341 |   },
342 |   {
343 |    "cell_type": "code",
344 |    "execution_count": 13,
345 |    "metadata": {},
346 |    "outputs": [
347 |     {
348 |      "name": "stdout",
349 |      "output_type": "stream",
350 |      "text": [
351 |       "69620.6890544\n",
352 |       "69901.4152052\n"
353 |      ]
354 |     }
355 |    ],
356 |    "source": [
357 |     "lr = LinearRegression()\n",
358 |     "\n",
359 |     "lr.fit(X_train, y_train)\n",
360 |     "\n",
361 |     "from sklearn.metrics.regression import mean_squared_error\n",
362 |     "\n",
363 |     "\n",
364 |     "y_predict = lr.predict(X_train)\n",
365 |     "\n",
366 |     "import numpy as np\n",
367 |     "\n",
368 |     "\n",
369 |     "print np.sqrt(mean_squared_error(y_train, y_predict))\n",
370 |     "\n",
371 |     "y_predict = lr.predict(X_test)\n",
372 |     "\n",
373 |     "\n",
374 |     "print np.sqrt(mean_squared_error(y_test, y_predict))"
375 |    ]
376 |   },
377 |   {
378 |    "cell_type": "code",
379 |    "execution_count": 14,
380 |    "metadata": {},
381 |    "outputs": [
382 |     {
383 |      "name": "stdout",
384 |      "output_type": "stream",
385 |      "text": [
386 |       "0.0\n",
387 |       "79017.4707616\n"
388 |      ]
389 |     }
390 |    ],
391 |    "source": [
392 |     "from sklearn.tree import DecisionTreeRegressor\n",
393 |     "\n",
394 |     "dtr = DecisionTreeRegressor()\n",
395 |     "\n",
396 |     "dtr.fit(X_train, y_train)\n",
397 |     "\n",
398 |     "y_predict = dtr.predict(X_train)\n",
399 |     "\n",
400 |     "print np.sqrt(mean_squared_error(y_train, y_predict))\n",
401 |     "\n",
402 |     "from sklearn.metrics.regression import mean_squared_error\n",
403 |     "\n",
404 |     "y_predict = dtr.predict(X_test)\n",
405 |     "\n",
406 |     "import numpy as np\n",
407 |     "print np.sqrt(mean_squared_error(y_test, y_predict))"
408 |    ]
409 |   },
410 |   {
411 |    "cell_type": "code",
412 |    "execution_count": null,
413 |    "metadata": {
414 |     "collapsed": true
415 |    },
416 |    "outputs": [],
417 |    "source": []
418 |   }
419 |  ],
420 |  "metadata": {
421 |   "kernelspec": {
422 |    "display_name": "Python 2",
423 |    "language": "python",
424 |    "name": "python2"
425 |   },
426 |   "language_info": {
427 |    "codemirror_mode": {
428 |     "name": "ipython",
429 |     "version": 2
430 |    },
431 |    "file_extension": ".py",
432 |    "mimetype": "text/x-python",
433 |    "name": "python",
434 |    "nbconvert_exporter": "python",
435 |    "pygments_lexer": "ipython2",
436 |    "version": "2.7.13"
437 |   }
438 |  },
439 |  "nbformat": 4,
440 |  "nbformat_minor": 2
441 | }
442 | 


--------------------------------------------------------------------------------
/MNIST_AUTOENCODER.ipynb:
--------------------------------------------------------------------------------
   1 | {
   2 |  "cells": [
   3 |   {
   4 |    "cell_type": "code",
   5 |    "execution_count": 1,
   6 |    "metadata": {
   7 |     "collapsed": true
   8 |    },
   9 |    "outputs": [],
  10 |    "source": [
  11 |     "import tensorflow as tf\n",
  12 |     "from tensorflow.examples.tutorials.mnist import input_data\n",
  13 |     "import numpy as np\n",
  14 |     "import matplotlib\n",
  15 |     "matplotlib.use('nbagg')\n",
  16 |     "\n",
  17 |     "import matplotlib.pyplot as plt"
  18 |    ]
  19 |   },
  20 |   {
  21 |    "cell_type": "code",
  22 |    "execution_count": 2,
  23 |    "metadata": {
  24 |     "collapsed": false
  25 |    },
  26 |    "outputs": [
  27 |     {
  28 |      "name": "stdout",
  29 |      "output_type": "stream",
  30 |      "text": [
  31 |       "Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
  32 |       "Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
  33 |       "Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
  34 |       "Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n"
  35 |      ]
  36 |     }
  37 |    ],
  38 |    "source": [
  39 |     "mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)"
  40 |    ]
  41 |   },
  42 |   {
  43 |    "cell_type": "code",
  44 |    "execution_count": 3,
  45 |    "metadata": {
  46 |     "collapsed": false
  47 |    },
  48 |    "outputs": [
  49 |     {
  50 |      "name": "stdout",
  51 |      "output_type": "stream",
  52 |      "text": [
  53 |       "(55000, 784) (55000, 10)\n"
  54 |      ]
  55 |     }
  56 |    ],
  57 |    "source": [
  58 |     "print mnist.train.images.shape, mnist.train.labels.shape"
  59 |    ]
  60 |   },
  61 |   {
  62 |    "cell_type": "code",
  63 |    "execution_count": 4,
  64 |    "metadata": {
  65 |     "collapsed": false
  66 |    },
  67 |    "outputs": [],
  68 |    "source": [
  69 |     "INPUT_UNITS = 784\n",
  70 |     "HIDDEN_UNITS = [256]\n",
  71 |     "ENCODED_UNITS = 128\n",
  72 |     "\n",
  73 |     "LEARNING_RATE = 1e-3\n",
  74 |     "TRAINING_ITER = 2000\n",
  75 |     "BATCH_SIZE = 100\n",
  76 |     "\n",
  77 |     "def init_weights(shape):\n",
  78 |     "    return tf.Variable(tf.truncated_normal(shape, stddev=0.1))\n",
  79 |     "\n",
  80 |     "def init_biases(shape):\n",
  81 |     "    return tf.Variable(tf.zeros(shape) + 0.1)\n",
  82 |     "\n",
  83 |     "def add_layer(x, weights, biases, activation_function = None):\n",
  84 |     "    scores = tf.matmul(x, weights) + biases\n",
  85 |     "    \n",
  86 |     "    if activation_function:\n",
  87 |     "        return activation_function(scores)\n",
  88 |     "    else:\n",
  89 |     "        return scores\n",
  90 |     "\n",
  91 |     "weights = {\n",
  92 |     "    'encoder_l1': init_weights([INPUT_UNITS, HIDDEN_UNITS[0]]), \n",
  93 |     "    'encoder_l2': init_weights([HIDDEN_UNITS[0], ENCODED_UNITS]),\n",
  94 |     "    'decoder_l1': init_weights([ENCODED_UNITS, HIDDEN_UNITS[0]]),\n",
  95 |     "    'decoder_l2': init_weights([HIDDEN_UNITS[0], INPUT_UNITS])\n",
  96 |     "}\n",
  97 |     "\n",
  98 |     "biases = {\n",
  99 |     "    'encoder_l1': init_biases([HIDDEN_UNITS[0]]), \n",
 100 |     "    'encoder_l2': init_biases([ENCODED_UNITS]),\n",
 101 |     "    'decoder_l1': init_biases(HIDDEN_UNITS[0]),\n",
 102 |     "    'decoder_l2': init_biases([INPUT_UNITS])\n",
 103 |     "}\n",
 104 |     "\n",
 105 |     "def encoder(x):\n",
 106 |     "    encoder_h1 = add_layer(x, weights['encoder_l1'], biases['encoder_l1'], tf.nn.relu)\n",
 107 |     "    return add_layer(encoder_h1, weights['encoder_l2'], biases['encoder_l2'], tf.nn.relu) \n",
 108 |     "\n",
 109 |     "def decoder(x):\n",
 110 |     "    decoder_h1 = add_layer(x, weights['decoder_l1'], biases['decoder_l1'], tf.nn.relu)\n",
 111 |     "    return add_layer(decoder_h1, weights['decoder_l2'], biases['decoder_l2'], tf.nn.relu)\n",
 112 |     "\n",
 113 |     "x = tf.placeholder(tf.float32, [None, 784])\n",
 114 |     "\n",
 115 |     "encoder_op = encoder(x)\n",
 116 |     "x_ = decoder(encoder_op)\n",
 117 |     "\n",
 118 |     "loss = tf.reduce_mean(tf.square(x - x_))\n",
 119 |     "\n",
 120 |     "train_step = tf.train.AdamOptimizer(LEARNING_RATE).minimize(loss)\n"
 121 |    ]
 122 |   },
 123 |   {
 124 |    "cell_type": "code",
 125 |    "execution_count": 5,
 126 |    "metadata": {
 127 |     "collapsed": false
 128 |    },
 129 |    "outputs": [
 130 |     {
 131 |      "name": "stdout",
 132 |      "output_type": "stream",
 133 |      "text": [
 134 |       "0.200395\n",
 135 |       "0.0587475\n",
 136 |       "0.039646\n",
 137 |       "0.0379133\n",
 138 |       "0.0267893\n",
 139 |       "0.026916\n",
 140 |       "0.0215286\n",
 141 |       "0.0191278\n",
 142 |       "0.0185347\n",
 143 |       "0.0168091\n",
 144 |       "0.0186432\n",
 145 |       "0.018459\n",
 146 |       "0.0159615\n",
 147 |       "0.0166012\n",
 148 |       "0.0161047\n",
 149 |       "0.0142662\n",
 150 |       "0.0149797\n",
 151 |       "0.0151406\n",
 152 |       "0.0143366\n",
 153 |       "0.0143979\n",
 154 |       "0.0138252\n",
 155 |       "0.0138254\n",
 156 |       "0.0128702\n",
 157 |       "0.0133265\n",
 158 |       "0.0130669\n",
 159 |       "0.0127066\n",
 160 |       "0.0140361\n",
 161 |       "0.012641\n",
 162 |       "0.013777\n",
 163 |       "0.0137268\n",
 164 |       "0.0132932\n",
 165 |       "0.0122787\n",
 166 |       "0.011369\n",
 167 |       "0.0119215\n",
 168 |       "0.0114884\n",
 169 |       "0.0122204\n",
 170 |       "0.0121677\n",
 171 |       "0.0114692\n",
 172 |       "0.0120117\n",
 173 |       "0.0117096\n"
 174 |      ]
 175 |     },
 176 |     {
 177 |      "data": {
 178 |       "application/javascript": [
 179 |        "/* Put everything inside the global mpl namespace */\n",
 180 |        "window.mpl = {};\n",
 181 |        "\n",
 182 |        "\n",
 183 |        "mpl.get_websocket_type = function() {\n",
 184 |        "    if (typeof(WebSocket) !== 'undefined') {\n",
 185 |        "        return WebSocket;\n",
 186 |        "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
 187 |        "        return MozWebSocket;\n",
 188 |        "    } else {\n",
 189 |        "        alert('Your browser does not have WebSocket support.' +\n",
 190 |        "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
 191 |        "              'Firefox 4 and 5 are also supported but you ' +\n",
 192 |        "              'have to enable WebSockets in about:config.');\n",
 193 |        "    };\n",
 194 |        "}\n",
 195 |        "\n",
 196 |        "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
 197 |        "    this.id = figure_id;\n",
 198 |        "\n",
 199 |        "    this.ws = websocket;\n",
 200 |        "\n",
 201 |        "    this.supports_binary = (this.ws.binaryType != undefined);\n",
 202 |        "\n",
 203 |        "    if (!this.supports_binary) {\n",
 204 |        "        var warnings = document.getElementById(\"mpl-warnings\");\n",
 205 |        "        if (warnings) {\n",
 206 |        "            warnings.style.display = 'block';\n",
 207 |        "            warnings.textContent = (\n",
 208 |        "                \"This browser does not support binary websocket messages. \" +\n",
 209 |        "                    \"Performance may be slow.\");\n",
 210 |        "        }\n",
 211 |        "    }\n",
 212 |        "\n",
 213 |        "    this.imageObj = new Image();\n",
 214 |        "\n",
 215 |        "    this.context = undefined;\n",
 216 |        "    this.message = undefined;\n",
 217 |        "    this.canvas = undefined;\n",
 218 |        "    this.rubberband_canvas = undefined;\n",
 219 |        "    this.rubberband_context = undefined;\n",
 220 |        "    this.format_dropdown = undefined;\n",
 221 |        "\n",
 222 |        "    this.image_mode = 'full';\n",
 223 |        "\n",
 224 |        "    this.root = $('<div/>');\n",
 225 |        "    this._root_extra_style(this.root)\n",
 226 |        "    this.root.attr('style', 'display: inline-block');\n",
 227 |        "\n",
 228 |        "    $(parent_element).append(this.root);\n",
 229 |        "\n",
 230 |        "    this._init_header(this);\n",
 231 |        "    this._init_canvas(this);\n",
 232 |        "    this._init_toolbar(this);\n",
 233 |        "\n",
 234 |        "    var fig = this;\n",
 235 |        "\n",
 236 |        "    this.waiting = false;\n",
 237 |        "\n",
 238 |        "    this.ws.onopen =  function () {\n",
 239 |        "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
 240 |        "            fig.send_message(\"send_image_mode\", {});\n",
 241 |        "            if (mpl.ratio != 1) {\n",
 242 |        "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
 243 |        "            }\n",
 244 |        "            fig.send_message(\"refresh\", {});\n",
 245 |        "        }\n",
 246 |        "\n",
 247 |        "    this.imageObj.onload = function() {\n",
 248 |        "            if (fig.image_mode == 'full') {\n",
 249 |        "                // Full images could contain transparency (where diff images\n",
 250 |        "                // almost always do), so we need to clear the canvas so that\n",
 251 |        "                // there is no ghosting.\n",
 252 |        "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
 253 |        "            }\n",
 254 |        "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
 255 |        "        };\n",
 256 |        "\n",
 257 |        "    this.imageObj.onunload = function() {\n",
 258 |        "        this.ws.close();\n",
 259 |        "    }\n",
 260 |        "\n",
 261 |        "    this.ws.onmessage = this._make_on_message_function(this);\n",
 262 |        "\n",
 263 |        "    this.ondownload = ondownload;\n",
 264 |        "}\n",
 265 |        "\n",
 266 |        "mpl.figure.prototype._init_header = function() {\n",
 267 |        "    var titlebar = $(\n",
 268 |        "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
 269 |        "        'ui-helper-clearfix\"/>');\n",
 270 |        "    var titletext = $(\n",
 271 |        "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
 272 |        "        'text-align: center; padding: 3px;\"/>');\n",
 273 |        "    titlebar.append(titletext)\n",
 274 |        "    this.root.append(titlebar);\n",
 275 |        "    this.header = titletext[0];\n",
 276 |        "}\n",
 277 |        "\n",
 278 |        "\n",
 279 |        "\n",
 280 |        "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
 281 |        "\n",
 282 |        "}\n",
 283 |        "\n",
 284 |        "\n",
 285 |        "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
 286 |        "\n",
 287 |        "}\n",
 288 |        "\n",
 289 |        "mpl.figure.prototype._init_canvas = function() {\n",
 290 |        "    var fig = this;\n",
 291 |        "\n",
 292 |        "    var canvas_div = $('<div/>');\n",
 293 |        "\n",
 294 |        "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
 295 |        "\n",
 296 |        "    function canvas_keyboard_event(event) {\n",
 297 |        "        return fig.key_event(event, event['data']);\n",
 298 |        "    }\n",
 299 |        "\n",
 300 |        "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
 301 |        "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
 302 |        "    this.canvas_div = canvas_div\n",
 303 |        "    this._canvas_extra_style(canvas_div)\n",
 304 |        "    this.root.append(canvas_div);\n",
 305 |        "\n",
 306 |        "    var canvas = $('<canvas/>');\n",
 307 |        "    canvas.addClass('mpl-canvas');\n",
 308 |        "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
 309 |        "\n",
 310 |        "    this.canvas = canvas[0];\n",
 311 |        "    this.context = canvas[0].getContext(\"2d\");\n",
 312 |        "\n",
 313 |        "    var backingStore = this.context.backingStorePixelRatio ||\n",
 314 |        "\tthis.context.webkitBackingStorePixelRatio ||\n",
 315 |        "\tthis.context.mozBackingStorePixelRatio ||\n",
 316 |        "\tthis.context.msBackingStorePixelRatio ||\n",
 317 |        "\tthis.context.oBackingStorePixelRatio ||\n",
 318 |        "\tthis.context.backingStorePixelRatio || 1;\n",
 319 |        "\n",
 320 |        "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
 321 |        "\n",
 322 |        "    var rubberband = $('<canvas/>');\n",
 323 |        "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
 324 |        "\n",
 325 |        "    var pass_mouse_events = true;\n",
 326 |        "\n",
 327 |        "    canvas_div.resizable({\n",
 328 |        "        start: function(event, ui) {\n",
 329 |        "            pass_mouse_events = false;\n",
 330 |        "        },\n",
 331 |        "        resize: function(event, ui) {\n",
 332 |        "            fig.request_resize(ui.size.width, ui.size.height);\n",
 333 |        "        },\n",
 334 |        "        stop: function(event, ui) {\n",
 335 |        "            pass_mouse_events = true;\n",
 336 |        "            fig.request_resize(ui.size.width, ui.size.height);\n",
 337 |        "        },\n",
 338 |        "    });\n",
 339 |        "\n",
 340 |        "    function mouse_event_fn(event) {\n",
 341 |        "        if (pass_mouse_events)\n",
 342 |        "            return fig.mouse_event(event, event['data']);\n",
 343 |        "    }\n",
 344 |        "\n",
 345 |        "    rubberband.mousedown('button_press', mouse_event_fn);\n",
 346 |        "    rubberband.mouseup('button_release', mouse_event_fn);\n",
 347 |        "    // Throttle sequential mouse events to 1 every 20ms.\n",
 348 |        "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
 349 |        "\n",
 350 |        "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
 351 |        "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
 352 |        "\n",
 353 |        "    canvas_div.on(\"wheel\", function (event) {\n",
 354 |        "        event = event.originalEvent;\n",
 355 |        "        event['data'] = 'scroll'\n",
 356 |        "        if (event.deltaY < 0) {\n",
 357 |        "            event.step = 1;\n",
 358 |        "        } else {\n",
 359 |        "            event.step = -1;\n",
 360 |        "        }\n",
 361 |        "        mouse_event_fn(event);\n",
 362 |        "    });\n",
 363 |        "\n",
 364 |        "    canvas_div.append(canvas);\n",
 365 |        "    canvas_div.append(rubberband);\n",
 366 |        "\n",
 367 |        "    this.rubberband = rubberband;\n",
 368 |        "    this.rubberband_canvas = rubberband[0];\n",
 369 |        "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
 370 |        "    this.rubberband_context.strokeStyle = \"#000000\";\n",
 371 |        "\n",
 372 |        "    this._resize_canvas = function(width, height) {\n",
 373 |        "        // Keep the size of the canvas, canvas container, and rubber band\n",
 374 |        "        // canvas in synch.\n",
 375 |        "        canvas_div.css('width', width)\n",
 376 |        "        canvas_div.css('height', height)\n",
 377 |        "\n",
 378 |        "        canvas.attr('width', width * mpl.ratio);\n",
 379 |        "        canvas.attr('height', height * mpl.ratio);\n",
 380 |        "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
 381 |        "\n",
 382 |        "        rubberband.attr('width', width);\n",
 383 |        "        rubberband.attr('height', height);\n",
 384 |        "    }\n",
 385 |        "\n",
 386 |        "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
 387 |        "    // upon first draw.\n",
 388 |        "    this._resize_canvas(600, 600);\n",
 389 |        "\n",
 390 |        "    // Disable right mouse context menu.\n",
 391 |        "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
 392 |        "        return false;\n",
 393 |        "    });\n",
 394 |        "\n",
 395 |        "    function set_focus () {\n",
 396 |        "        canvas.focus();\n",
 397 |        "        canvas_div.focus();\n",
 398 |        "    }\n",
 399 |        "\n",
 400 |        "    window.setTimeout(set_focus, 100);\n",
 401 |        "}\n",
 402 |        "\n",
 403 |        "mpl.figure.prototype._init_toolbar = function() {\n",
 404 |        "    var fig = this;\n",
 405 |        "\n",
 406 |        "    var nav_element = $('<div/>')\n",
 407 |        "    nav_element.attr('style', 'width: 100%');\n",
 408 |        "    this.root.append(nav_element);\n",
 409 |        "\n",
 410 |        "    // Define a callback function for later on.\n",
 411 |        "    function toolbar_event(event) {\n",
 412 |        "        return fig.toolbar_button_onclick(event['data']);\n",
 413 |        "    }\n",
 414 |        "    function toolbar_mouse_event(event) {\n",
 415 |        "        return fig.toolbar_button_onmouseover(event['data']);\n",
 416 |        "    }\n",
 417 |        "\n",
 418 |        "    for(var toolbar_ind in mpl.toolbar_items) {\n",
 419 |        "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
 420 |        "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
 421 |        "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
 422 |        "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
 423 |        "\n",
 424 |        "        if (!name) {\n",
 425 |        "            // put a spacer in here.\n",
 426 |        "            continue;\n",
 427 |        "        }\n",
 428 |        "        var button = $('<button/>');\n",
 429 |        "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
 430 |        "                        'ui-button-icon-only');\n",
 431 |        "        button.attr('role', 'button');\n",
 432 |        "        button.attr('aria-disabled', 'false');\n",
 433 |        "        button.click(method_name, toolbar_event);\n",
 434 |        "        button.mouseover(tooltip, toolbar_mouse_event);\n",
 435 |        "\n",
 436 |        "        var icon_img = $('<span/>');\n",
 437 |        "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
 438 |        "        icon_img.addClass(image);\n",
 439 |        "        icon_img.addClass('ui-corner-all');\n",
 440 |        "\n",
 441 |        "        var tooltip_span = $('<span/>');\n",
 442 |        "        tooltip_span.addClass('ui-button-text');\n",
 443 |        "        tooltip_span.html(tooltip);\n",
 444 |        "\n",
 445 |        "        button.append(icon_img);\n",
 446 |        "        button.append(tooltip_span);\n",
 447 |        "\n",
 448 |        "        nav_element.append(button);\n",
 449 |        "    }\n",
 450 |        "\n",
 451 |        "    var fmt_picker_span = $('<span/>');\n",
 452 |        "\n",
 453 |        "    var fmt_picker = $('<select/>');\n",
 454 |        "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
 455 |        "    fmt_picker_span.append(fmt_picker);\n",
 456 |        "    nav_element.append(fmt_picker_span);\n",
 457 |        "    this.format_dropdown = fmt_picker[0];\n",
 458 |        "\n",
 459 |        "    for (var ind in mpl.extensions) {\n",
 460 |        "        var fmt = mpl.extensions[ind];\n",
 461 |        "        var option = $(\n",
 462 |        "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
 463 |        "        fmt_picker.append(option)\n",
 464 |        "    }\n",
 465 |        "\n",
 466 |        "    // Add hover states to the ui-buttons\n",
 467 |        "    $( \".ui-button\" ).hover(\n",
 468 |        "        function() { $(this).addClass(\"ui-state-hover\");},\n",
 469 |        "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
 470 |        "    );\n",
 471 |        "\n",
 472 |        "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
 473 |        "    nav_element.append(status_bar);\n",
 474 |        "    this.message = status_bar[0];\n",
 475 |        "}\n",
 476 |        "\n",
 477 |        "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
 478 |        "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
 479 |        "    // which will in turn request a refresh of the image.\n",
 480 |        "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
 481 |        "}\n",
 482 |        "\n",
 483 |        "mpl.figure.prototype.send_message = function(type, properties) {\n",
 484 |        "    properties['type'] = type;\n",
 485 |        "    properties['figure_id'] = this.id;\n",
 486 |        "    this.ws.send(JSON.stringify(properties));\n",
 487 |        "}\n",
 488 |        "\n",
 489 |        "mpl.figure.prototype.send_draw_message = function() {\n",
 490 |        "    if (!this.waiting) {\n",
 491 |        "        this.waiting = true;\n",
 492 |        "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
 493 |        "    }\n",
 494 |        "}\n",
 495 |        "\n",
 496 |        "\n",
 497 |        "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
 498 |        "    var format_dropdown = fig.format_dropdown;\n",
 499 |        "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
 500 |        "    fig.ondownload(fig, format);\n",
 501 |        "}\n",
 502 |        "\n",
 503 |        "\n",
 504 |        "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
 505 |        "    var size = msg['size'];\n",
 506 |        "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
 507 |        "        fig._resize_canvas(size[0], size[1]);\n",
 508 |        "        fig.send_message(\"refresh\", {});\n",
 509 |        "    };\n",
 510 |        "}\n",
 511 |        "\n",
 512 |        "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
 513 |        "    var x0 = msg['x0'] / mpl.ratio;\n",
 514 |        "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
 515 |        "    var x1 = msg['x1'] / mpl.ratio;\n",
 516 |        "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
 517 |        "    x0 = Math.floor(x0) + 0.5;\n",
 518 |        "    y0 = Math.floor(y0) + 0.5;\n",
 519 |        "    x1 = Math.floor(x1) + 0.5;\n",
 520 |        "    y1 = Math.floor(y1) + 0.5;\n",
 521 |        "    var min_x = Math.min(x0, x1);\n",
 522 |        "    var min_y = Math.min(y0, y1);\n",
 523 |        "    var width = Math.abs(x1 - x0);\n",
 524 |        "    var height = Math.abs(y1 - y0);\n",
 525 |        "\n",
 526 |        "    fig.rubberband_context.clearRect(\n",
 527 |        "        0, 0, fig.canvas.width, fig.canvas.height);\n",
 528 |        "\n",
 529 |        "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
 530 |        "}\n",
 531 |        "\n",
 532 |        "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
 533 |        "    // Updates the figure title.\n",
 534 |        "    fig.header.textContent = msg['label'];\n",
 535 |        "}\n",
 536 |        "\n",
 537 |        "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
 538 |        "    var cursor = msg['cursor'];\n",
 539 |        "    switch(cursor)\n",
 540 |        "    {\n",
 541 |        "    case 0:\n",
 542 |        "        cursor = 'pointer';\n",
 543 |        "        break;\n",
 544 |        "    case 1:\n",
 545 |        "        cursor = 'default';\n",
 546 |        "        break;\n",
 547 |        "    case 2:\n",
 548 |        "        cursor = 'crosshair';\n",
 549 |        "        break;\n",
 550 |        "    case 3:\n",
 551 |        "        cursor = 'move';\n",
 552 |        "        break;\n",
 553 |        "    }\n",
 554 |        "    fig.rubberband_canvas.style.cursor = cursor;\n",
 555 |        "}\n",
 556 |        "\n",
 557 |        "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
 558 |        "    fig.message.textContent = msg['message'];\n",
 559 |        "}\n",
 560 |        "\n",
 561 |        "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
 562 |        "    // Request the server to send over a new figure.\n",
 563 |        "    fig.send_draw_message();\n",
 564 |        "}\n",
 565 |        "\n",
 566 |        "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
 567 |        "    fig.image_mode = msg['mode'];\n",
 568 |        "}\n",
 569 |        "\n",
 570 |        "mpl.figure.prototype.updated_canvas_event = function() {\n",
 571 |        "    // Called whenever the canvas gets updated.\n",
 572 |        "    this.send_message(\"ack\", {});\n",
 573 |        "}\n",
 574 |        "\n",
 575 |        "// A function to construct a web socket function for onmessage handling.\n",
 576 |        "// Called in the figure constructor.\n",
 577 |        "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
 578 |        "    return function socket_on_message(evt) {\n",
 579 |        "        if (evt.data instanceof Blob) {\n",
 580 |        "            /* FIXME: We get \"Resource interpreted as Image but\n",
 581 |        "             * transferred with MIME type text/plain:\" errors on\n",
 582 |        "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
 583 |        "             * to be part of the websocket stream */\n",
 584 |        "            evt.data.type = \"image/png\";\n",
 585 |        "\n",
 586 |        "            /* Free the memory for the previous frames */\n",
 587 |        "            if (fig.imageObj.src) {\n",
 588 |        "                (window.URL || window.webkitURL).revokeObjectURL(\n",
 589 |        "                    fig.imageObj.src);\n",
 590 |        "            }\n",
 591 |        "\n",
 592 |        "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
 593 |        "                evt.data);\n",
 594 |        "            fig.updated_canvas_event();\n",
 595 |        "            fig.waiting = false;\n",
 596 |        "            return;\n",
 597 |        "        }\n",
 598 |        "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
 599 |        "            fig.imageObj.src = evt.data;\n",
 600 |        "            fig.updated_canvas_event();\n",
 601 |        "            fig.waiting = false;\n",
 602 |        "            return;\n",
 603 |        "        }\n",
 604 |        "\n",
 605 |        "        var msg = JSON.parse(evt.data);\n",
 606 |        "        var msg_type = msg['type'];\n",
 607 |        "\n",
 608 |        "        // Call the  \"handle_{type}\" callback, which takes\n",
 609 |        "        // the figure and JSON message as its only arguments.\n",
 610 |        "        try {\n",
 611 |        "            var callback = fig[\"handle_\" + msg_type];\n",
 612 |        "        } catch (e) {\n",
 613 |        "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
 614 |        "            return;\n",
 615 |        "        }\n",
 616 |        "\n",
 617 |        "        if (callback) {\n",
 618 |        "            try {\n",
 619 |        "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
 620 |        "                callback(fig, msg);\n",
 621 |        "            } catch (e) {\n",
 622 |        "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
 623 |        "            }\n",
 624 |        "        }\n",
 625 |        "    };\n",
 626 |        "}\n",
 627 |        "\n",
 628 |        "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
 629 |        "mpl.findpos = function(e) {\n",
 630 |        "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
 631 |        "    var targ;\n",
 632 |        "    if (!e)\n",
 633 |        "        e = window.event;\n",
 634 |        "    if (e.target)\n",
 635 |        "        targ = e.target;\n",
 636 |        "    else if (e.srcElement)\n",
 637 |        "        targ = e.srcElement;\n",
 638 |        "    if (targ.nodeType == 3) // defeat Safari bug\n",
 639 |        "        targ = targ.parentNode;\n",
 640 |        "\n",
 641 |        "    // jQuery normalizes the pageX and pageY\n",
 642 |        "    // pageX,Y are the mouse positions relative to the document\n",
 643 |        "    // offset() returns the position of the element relative to the document\n",
 644 |        "    var x = e.pageX - $(targ).offset().left;\n",
 645 |        "    var y = e.pageY - $(targ).offset().top;\n",
 646 |        "\n",
 647 |        "    return {\"x\": x, \"y\": y};\n",
 648 |        "};\n",
 649 |        "\n",
 650 |        "/*\n",
 651 |        " * return a copy of an object with only non-object keys\n",
 652 |        " * we need this to avoid circular references\n",
 653 |        " * http://stackoverflow.com/a/24161582/3208463\n",
 654 |        " */\n",
 655 |        "function simpleKeys (original) {\n",
 656 |        "  return Object.keys(original).reduce(function (obj, key) {\n",
 657 |        "    if (typeof original[key] !== 'object')\n",
 658 |        "        obj[key] = original[key]\n",
 659 |        "    return obj;\n",
 660 |        "  }, {});\n",
 661 |        "}\n",
 662 |        "\n",
 663 |        "mpl.figure.prototype.mouse_event = function(event, name) {\n",
 664 |        "    var canvas_pos = mpl.findpos(event)\n",
 665 |        "\n",
 666 |        "    if (name === 'button_press')\n",
 667 |        "    {\n",
 668 |        "        this.canvas.focus();\n",
 669 |        "        this.canvas_div.focus();\n",
 670 |        "    }\n",
 671 |        "\n",
 672 |        "    var x = canvas_pos.x * mpl.ratio;\n",
 673 |        "    var y = canvas_pos.y * mpl.ratio;\n",
 674 |        "\n",
 675 |        "    this.send_message(name, {x: x, y: y, button: event.button,\n",
 676 |        "                             step: event.step,\n",
 677 |        "                             guiEvent: simpleKeys(event)});\n",
 678 |        "\n",
 679 |        "    /* This prevents the web browser from automatically changing to\n",
 680 |        "     * the text insertion cursor when the button is pressed.  We want\n",
 681 |        "     * to control all of the cursor setting manually through the\n",
 682 |        "     * 'cursor' event from matplotlib */\n",
 683 |        "    event.preventDefault();\n",
 684 |        "    return false;\n",
 685 |        "}\n",
 686 |        "\n",
 687 |        "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
 688 |        "    // Handle any extra behaviour associated with a key event\n",
 689 |        "}\n",
 690 |        "\n",
 691 |        "mpl.figure.prototype.key_event = function(event, name) {\n",
 692 |        "\n",
 693 |        "    // Prevent repeat events\n",
 694 |        "    if (name == 'key_press')\n",
 695 |        "    {\n",
 696 |        "        if (event.which === this._key)\n",
 697 |        "            return;\n",
 698 |        "        else\n",
 699 |        "            this._key = event.which;\n",
 700 |        "    }\n",
 701 |        "    if (name == 'key_release')\n",
 702 |        "        this._key = null;\n",
 703 |        "\n",
 704 |        "    var value = '';\n",
 705 |        "    if (event.ctrlKey && event.which != 17)\n",
 706 |        "        value += \"ctrl+\";\n",
 707 |        "    if (event.altKey && event.which != 18)\n",
 708 |        "        value += \"alt+\";\n",
 709 |        "    if (event.shiftKey && event.which != 16)\n",
 710 |        "        value += \"shift+\";\n",
 711 |        "\n",
 712 |        "    value += 'k';\n",
 713 |        "    value += event.which.toString();\n",
 714 |        "\n",
 715 |        "    this._key_event_extra(event, name);\n",
 716 |        "\n",
 717 |        "    this.send_message(name, {key: value,\n",
 718 |        "                             guiEvent: simpleKeys(event)});\n",
 719 |        "    return false;\n",
 720 |        "}\n",
 721 |        "\n",
 722 |        "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
 723 |        "    if (name == 'download') {\n",
 724 |        "        this.handle_save(this, null);\n",
 725 |        "    } else {\n",
 726 |        "        this.send_message(\"toolbar_button\", {name: name});\n",
 727 |        "    }\n",
 728 |        "};\n",
 729 |        "\n",
 730 |        "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
 731 |        "    this.message.textContent = tooltip;\n",
 732 |        "};\n",
 733 |        "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to  previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
 734 |        "\n",
 735 |        "mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
 736 |        "\n",
 737 |        "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
 738 |        "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
 739 |        "    // object with the appropriate methods. Currently this is a non binary\n",
 740 |        "    // socket, so there is still some room for performance tuning.\n",
 741 |        "    var ws = {};\n",
 742 |        "\n",
 743 |        "    ws.close = function() {\n",
 744 |        "        comm.close()\n",
 745 |        "    };\n",
 746 |        "    ws.send = function(m) {\n",
 747 |        "        //console.log('sending', m);\n",
 748 |        "        comm.send(m);\n",
 749 |        "    };\n",
 750 |        "    // Register the callback with on_msg.\n",
 751 |        "    comm.on_msg(function(msg) {\n",
 752 |        "        //console.log('receiving', msg['content']['data'], msg);\n",
 753 |        "        // Pass the mpl event to the overriden (by mpl) onmessage function.\n",
 754 |        "        ws.onmessage(msg['content']['data'])\n",
 755 |        "    });\n",
 756 |        "    return ws;\n",
 757 |        "}\n",
 758 |        "\n",
 759 |        "mpl.mpl_figure_comm = function(comm, msg) {\n",
 760 |        "    // This is the function which gets called when the mpl process\n",
 761 |        "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
 762 |        "\n",
 763 |        "    var id = msg.content.data.id;\n",
 764 |        "    // Get hold of the div created by the display call when the Comm\n",
 765 |        "    // socket was opened in Python.\n",
 766 |        "    var element = $(\"#\" + id);\n",
 767 |        "    var ws_proxy = comm_websocket_adapter(comm)\n",
 768 |        "\n",
 769 |        "    function ondownload(figure, format) {\n",
 770 |        "        window.open(figure.imageObj.src);\n",
 771 |        "    }\n",
 772 |        "\n",
 773 |        "    var fig = new mpl.figure(id, ws_proxy,\n",
 774 |        "                           ondownload,\n",
 775 |        "                           element.get(0));\n",
 776 |        "\n",
 777 |        "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
 778 |        "    // web socket which is closed, not our websocket->open comm proxy.\n",
 779 |        "    ws_proxy.onopen();\n",
 780 |        "\n",
 781 |        "    fig.parent_element = element.get(0);\n",
 782 |        "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
 783 |        "    if (!fig.cell_info) {\n",
 784 |        "        console.error(\"Failed to find cell for figure\", id, fig);\n",
 785 |        "        return;\n",
 786 |        "    }\n",
 787 |        "\n",
 788 |        "    var output_index = fig.cell_info[2]\n",
 789 |        "    var cell = fig.cell_info[0];\n",
 790 |        "\n",
 791 |        "};\n",
 792 |        "\n",
 793 |        "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
 794 |        "    var width = fig.canvas.width/mpl.ratio\n",
 795 |        "    fig.root.unbind('remove')\n",
 796 |        "\n",
 797 |        "    // Update the output cell to use the data from the current canvas.\n",
 798 |        "    fig.push_to_output();\n",
 799 |        "    var dataURL = fig.canvas.toDataURL();\n",
 800 |        "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
 801 |        "    // the notebook keyboard shortcuts fail.\n",
 802 |        "    IPython.keyboard_manager.enable()\n",
 803 |        "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
 804 |        "    fig.close_ws(fig, msg);\n",
 805 |        "}\n",
 806 |        "\n",
 807 |        "mpl.figure.prototype.close_ws = function(fig, msg){\n",
 808 |        "    fig.send_message('closing', msg);\n",
 809 |        "    // fig.ws.close()\n",
 810 |        "}\n",
 811 |        "\n",
 812 |        "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
 813 |        "    // Turn the data on the canvas into data in the output cell.\n",
 814 |        "    var width = this.canvas.width/mpl.ratio\n",
 815 |        "    var dataURL = this.canvas.toDataURL();\n",
 816 |        "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
 817 |        "}\n",
 818 |        "\n",
 819 |        "mpl.figure.prototype.updated_canvas_event = function() {\n",
 820 |        "    // Tell IPython that the notebook contents must change.\n",
 821 |        "    IPython.notebook.set_dirty(true);\n",
 822 |        "    this.send_message(\"ack\", {});\n",
 823 |        "    var fig = this;\n",
 824 |        "    // Wait a second, then push the new image to the DOM so\n",
 825 |        "    // that it is saved nicely (might be nice to debounce this).\n",
 826 |        "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
 827 |        "}\n",
 828 |        "\n",
 829 |        "mpl.figure.prototype._init_toolbar = function() {\n",
 830 |        "    var fig = this;\n",
 831 |        "\n",
 832 |        "    var nav_element = $('<div/>')\n",
 833 |        "    nav_element.attr('style', 'width: 100%');\n",
 834 |        "    this.root.append(nav_element);\n",
 835 |        "\n",
 836 |        "    // Define a callback function for later on.\n",
 837 |        "    function toolbar_event(event) {\n",
 838 |        "        return fig.toolbar_button_onclick(event['data']);\n",
 839 |        "    }\n",
 840 |        "    function toolbar_mouse_event(event) {\n",
 841 |        "        return fig.toolbar_button_onmouseover(event['data']);\n",
 842 |        "    }\n",
 843 |        "\n",
 844 |        "    for(var toolbar_ind in mpl.toolbar_items){\n",
 845 |        "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
 846 |        "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
 847 |        "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
 848 |        "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
 849 |        "\n",
 850 |        "        if (!name) { continue; };\n",
 851 |        "\n",
 852 |        "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
 853 |        "        button.click(method_name, toolbar_event);\n",
 854 |        "        button.mouseover(tooltip, toolbar_mouse_event);\n",
 855 |        "        nav_element.append(button);\n",
 856 |        "    }\n",
 857 |        "\n",
 858 |        "    // Add the status bar.\n",
 859 |        "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
 860 |        "    nav_element.append(status_bar);\n",
 861 |        "    this.message = status_bar[0];\n",
 862 |        "\n",
 863 |        "    // Add the close button to the window.\n",
 864 |        "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
 865 |        "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
 866 |        "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
 867 |        "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
 868 |        "    buttongrp.append(button);\n",
 869 |        "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
 870 |        "    titlebar.prepend(buttongrp);\n",
 871 |        "}\n",
 872 |        "\n",
 873 |        "mpl.figure.prototype._root_extra_style = function(el){\n",
 874 |        "    var fig = this\n",
 875 |        "    el.on(\"remove\", function(){\n",
 876 |        "\tfig.close_ws(fig, {});\n",
 877 |        "    });\n",
 878 |        "}\n",
 879 |        "\n",
 880 |        "mpl.figure.prototype._canvas_extra_style = function(el){\n",
 881 |        "    // this is important to make the div 'focusable\n",
 882 |        "    el.attr('tabindex', 0)\n",
 883 |        "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
 884 |        "    // off when our div gets focus\n",
 885 |        "\n",
 886 |        "    // location in version 3\n",
 887 |        "    if (IPython.notebook.keyboard_manager) {\n",
 888 |        "        IPython.notebook.keyboard_manager.register_events(el);\n",
 889 |        "    }\n",
 890 |        "    else {\n",
 891 |        "        // location in version 2\n",
 892 |        "        IPython.keyboard_manager.register_events(el);\n",
 893 |        "    }\n",
 894 |        "\n",
 895 |        "}\n",
 896 |        "\n",
 897 |        "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
 898 |        "    var manager = IPython.notebook.keyboard_manager;\n",
 899 |        "    if (!manager)\n",
 900 |        "        manager = IPython.keyboard_manager;\n",
 901 |        "\n",
 902 |        "    // Check for shift+enter\n",
 903 |        "    if (event.shiftKey && event.which == 13) {\n",
 904 |        "        this.canvas_div.blur();\n",
 905 |        "        // select the cell after this one\n",
 906 |        "        var index = IPython.notebook.find_cell_index(this.cell_info[0]);\n",
 907 |        "        IPython.notebook.select(index + 1);\n",
 908 |        "    }\n",
 909 |        "}\n",
 910 |        "\n",
 911 |        "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
 912 |        "    fig.ondownload(fig, null);\n",
 913 |        "}\n",
 914 |        "\n",
 915 |        "\n",
 916 |        "mpl.find_output_cell = function(html_output) {\n",
 917 |        "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
 918 |        "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
 919 |        "    // IPython event is triggered only after the cells have been serialised, which for\n",
 920 |        "    // our purposes (turning an active figure into a static one), is too late.\n",
 921 |        "    var cells = IPython.notebook.get_cells();\n",
 922 |        "    var ncells = cells.length;\n",
 923 |        "    for (var i=0; i<ncells; i++) {\n",
 924 |        "        var cell = cells[i];\n",
 925 |        "        if (cell.cell_type === 'code'){\n",
 926 |        "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
 927 |        "                var data = cell.output_area.outputs[j];\n",
 928 |        "                if (data.data) {\n",
 929 |        "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
 930 |        "                    data = data.data;\n",
 931 |        "                }\n",
 932 |        "                if (data['text/html'] == html_output) {\n",
 933 |        "                    return [cell, data, j];\n",
 934 |        "                }\n",
 935 |        "            }\n",
 936 |        "        }\n",
 937 |        "    }\n",
 938 |        "}\n",
 939 |        "\n",
 940 |        "// Register the function which deals with the matplotlib target/channel.\n",
 941 |        "// The kernel may be null if the page has been refreshed.\n",
 942 |        "if (IPython.notebook.kernel != null) {\n",
 943 |        "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
 944 |        "}\n"
 945 |       ],
 946 |       "text/plain": [
 947 |        "<IPython.core.display.Javascript object>"
 948 |       ]
 949 |      },
 950 |      "metadata": {},
 951 |      "output_type": "display_data"
 952 |     },
 953 |     {
 954 |      "data": {
 955 |       "text/html": [
 956 |        "<img src=\"data:image/png;base64,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\" width=\"1000\">"
 957 |       ],
 958 |       "text/plain": [
 959 |        "<IPython.core.display.HTML object>"
 960 |       ]
 961 |      },
 962 |      "metadata": {},
 963 |      "output_type": "display_data"
 964 |     }
 965 |    ],
 966 |    "source": [
 967 |     "init = tf.global_variables_initializer()\n",
 968 |     "\n",
 969 |     "with tf.Session() as sess:\n",
 970 |     "    sess.run(init)\n",
 971 |     "    for i in range(TRAINING_ITER):\n",
 972 |     "        batch_xs, _ = mnist.train.next_batch(BATCH_SIZE)\n",
 973 |     "        sess.run(train_step, feed_dict = {x: batch_xs})\n",
 974 |     "        if i % 50 == 0:\n",
 975 |     "            print sess.run(loss, feed_dict = {x: batch_xs})\n",
 976 |     "    encode_decode = sess.run(x_, feed_dict={x: mnist.test.images[:10]})\n",
 977 |     "    # Compare original images with their reconstructions\n",
 978 |     "    f, a = plt.subplots(2, 10, figsize=(10, 2))\n",
 979 |     "    for i in range(10):\n",
 980 |     "        a[0][i].imshow(np.reshape(mnist.test.images[i], (28, 28)))\n",
 981 |     "        a[1][i].imshow(np.reshape(encode_decode[i], (28, 28)))\n",
 982 |     "    plt.show()"
 983 |    ]
 984 |   },
 985 |   {
 986 |    "cell_type": "code",
 987 |    "execution_count": null,
 988 |    "metadata": {
 989 |     "collapsed": false
 990 |    },
 991 |    "outputs": [],
 992 |    "source": []
 993 |   },
 994 |   {
 995 |    "cell_type": "code",
 996 |    "execution_count": null,
 997 |    "metadata": {
 998 |     "collapsed": true
 999 |    },
1000 |    "outputs": [],
1001 |    "source": []
1002 |   },
1003 |   {
1004 |    "cell_type": "code",
1005 |    "execution_count": null,
1006 |    "metadata": {
1007 |     "collapsed": true
1008 |    },
1009 |    "outputs": [],
1010 |    "source": []
1011 |   }
1012 |  ],
1013 |  "metadata": {
1014 |   "kernelspec": {
1015 |    "display_name": "Python 2",
1016 |    "language": "python",
1017 |    "name": "python2"
1018 |   },
1019 |   "language_info": {
1020 |    "codemirror_mode": {
1021 |     "name": "ipython",
1022 |     "version": 2
1023 |    },
1024 |    "file_extension": ".py",
1025 |    "mimetype": "text/x-python",
1026 |    "name": "python",
1027 |    "nbconvert_exporter": "python",
1028 |    "pygments_lexer": "ipython2",
1029 |    "version": "2.7.11"
1030 |   }
1031 |  },
1032 |  "nbformat": 4,
1033 |  "nbformat_minor": 0
1034 | }
1035 | 


--------------------------------------------------------------------------------
/MNIST_BILSTM.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 1,
  6 |    "metadata": {
  7 |     "collapsed": true
  8 |    },
  9 |    "outputs": [],
 10 |    "source": [
 11 |     "from tensorflow.examples.tutorials.mnist import input_data\n",
 12 |     "import tensorflow as tf"
 13 |    ]
 14 |   },
 15 |   {
 16 |    "cell_type": "code",
 17 |    "execution_count": 2,
 18 |    "metadata": {
 19 |     "collapsed": false
 20 |    },
 21 |    "outputs": [
 22 |     {
 23 |      "name": "stdout",
 24 |      "output_type": "stream",
 25 |      "text": [
 26 |       "Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
 27 |       "Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
 28 |       "Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
 29 |       "Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n"
 30 |      ]
 31 |     }
 32 |    ],
 33 |    "source": [
 34 |     "mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)"
 35 |    ]
 36 |   },
 37 |   {
 38 |    "cell_type": "code",
 39 |    "execution_count": 3,
 40 |    "metadata": {
 41 |     "collapsed": false
 42 |    },
 43 |    "outputs": [],
 44 |    "source": [
 45 |     "HIDDEN_UNITS = 256\n",
 46 |     "N_CLASSES = 10\n",
 47 |     "LEARNING_RATE = 1e-2\n",
 48 |     "TRAINING_ITER = 1000\n",
 49 |     "BATCH_SIZE = 100\n",
 50 |     "\n",
 51 |     "def BiRNN(x, weights, biases):\n",
 52 |     "    x = tf.transpose(x, [1,0,2])\n",
 53 |     "    x = tf.reshape(x, [-1, 28])\n",
 54 |     "    x = tf.split(x, 28)\n",
 55 |     "    \n",
 56 |     "    lstm_fw_cell = tf.contrib.rnn.BasicLSTMCell(num_units=HIDDEN_UNITS)\n",
 57 |     "    lstm_bw_cell = tf.contrib.rnn.BasicLSTMCell(num_units=HIDDEN_UNITS)\n",
 58 |     "    \n",
 59 |     "    outputs, _, _ = tf.contrib.rnn.static_bidirectional_rnn(lstm_fw_cell, lstm_bw_cell, x, dtype=tf.float32)\n",
 60 |     "    \n",
 61 |     "    return tf.matmul(outputs[-1], weights) + biases\n",
 62 |     "\n",
 63 |     "\n",
 64 |     "x = tf.placeholder(tf.float32, [None, 784])\n",
 65 |     "y = tf.placeholder(tf.float32, [None, 10])\n",
 66 |     "\n",
 67 |     "x_2d = tf.reshape(x, [-1, 28, 28])\n",
 68 |     "\n",
 69 |     "weights = {'output': tf.Variable(tf.random_normal([2 * HIDDEN_UNITS, N_CLASSES]))}\n",
 70 |     "biases = {'output': tf.Variable(tf.zeros([N_CLASSES]) + 0.1)}\n",
 71 |     "\n",
 72 |     "predictions = BiRNN(x_2d, weights['output'], biases['output'])\n",
 73 |     "\n",
 74 |     "loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=predictions))\n",
 75 |     "train_step = tf.train.AdamOptimizer(LEARNING_RATE).minimize(loss)\n",
 76 |     "\n",
 77 |     "accuracy = tf.reduce_mean(tf.cast(tf.equal(tf.arg_max(y, 1), tf.arg_max(predictions, 1)), tf.float32))\n",
 78 |     "\n",
 79 |     "\n",
 80 |     "init = tf.global_variables_initializer()\n"
 81 |    ]
 82 |   },
 83 |   {
 84 |    "cell_type": "code",
 85 |    "execution_count": 5,
 86 |    "metadata": {
 87 |     "collapsed": false
 88 |    },
 89 |    "outputs": [
 90 |     {
 91 |      "name": "stdout",
 92 |      "output_type": "stream",
 93 |      "text": [
 94 |       "0.15\n",
 95 |       "0.73\n",
 96 |       "0.84\n",
 97 |       "0.92\n",
 98 |       "0.95\n",
 99 |       "0.91\n",
100 |       "0.95\n",
101 |       "0.98\n",
102 |       "0.99\n",
103 |       "0.97\n",
104 |       "0.97\n",
105 |       "0.99\n",
106 |       "0.99\n",
107 |       "0.98\n",
108 |       "0.99\n",
109 |       "0.99\n"
110 |      ]
111 |     },
112 |     {
113 |      "ename": "KeyboardInterrupt",
114 |      "evalue": "",
115 |      "output_type": "error",
116 |      "traceback": [
117 |       "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
118 |       "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
119 |       "\u001b[0;32m<ipython-input-5-e01b19a1e24e>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      3\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mTRAINING_ITER\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m         \u001b[0mbatch_xs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_ys\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmnist\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnext_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mBATCH_SIZE\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m         \u001b[0msess\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_step\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mbatch_xs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mbatch_ys\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      6\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mi\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0;36m50\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m             \u001b[0;32mprint\u001b[0m \u001b[0msess\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0maccuracy\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mbatch_xs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mbatch_ys\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
120 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    765\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    766\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 767\u001b[0;31m                          run_metadata_ptr)\n\u001b[0m\u001b[1;32m    768\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    769\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
121 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    963\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    964\u001b[0m       results = self._do_run(handle, final_targets, final_fetches,\n\u001b[0;32m--> 965\u001b[0;31m                              feed_dict_string, options, run_metadata)\n\u001b[0m\u001b[1;32m    966\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    967\u001b[0m       \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
122 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_do_run\u001b[0;34m(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1013\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1014\u001b[0m       return self._do_call(_run_fn, self._session, feed_dict, fetch_list,\n\u001b[0;32m-> 1015\u001b[0;31m                            target_list, options, run_metadata)\n\u001b[0m\u001b[1;32m   1016\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1017\u001b[0m       return self._do_call(_prun_fn, self._session, handle, feed_dict,\n",
123 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_do_call\u001b[0;34m(self, fn, *args)\u001b[0m\n\u001b[1;32m   1020\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1021\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1022\u001b[0;31m       \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1023\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0merrors\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mOpError\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1024\u001b[0m       \u001b[0mmessage\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcompat\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mas_text\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmessage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
124 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_run_fn\u001b[0;34m(session, feed_dict, fetch_list, target_list, options, run_metadata)\u001b[0m\n\u001b[1;32m   1002\u001b[0m         return tf_session.TF_Run(session, options,\n\u001b[1;32m   1003\u001b[0m                                  \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtarget_list\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1004\u001b[0;31m                                  status, run_metadata)\n\u001b[0m\u001b[1;32m   1005\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1006\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_prun_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msession\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
125 |       "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
126 |      ]
127 |     }
128 |    ],
129 |    "source": [
130 |     "with tf.Session() as sess:\n",
131 |     "    sess.run(init)\n",
132 |     "    for i in range(TRAINING_ITER):\n",
133 |     "        batch_xs, batch_ys = mnist.train.next_batch(BATCH_SIZE)\n",
134 |     "        sess.run(train_step, feed_dict = {x: batch_xs, y: batch_ys})\n",
135 |     "        if i % 50 == 0:\n",
136 |     "            print sess.run(accuracy, feed_dict = {x: batch_xs, y: batch_ys})\n",
137 |     "            print sess.run(accuracy, feed_dict = {x: mnist.validation.images, y: mnist.validation.labels})"
138 |    ]
139 |   },
140 |   {
141 |    "cell_type": "code",
142 |    "execution_count": null,
143 |    "metadata": {
144 |     "collapsed": true
145 |    },
146 |    "outputs": [],
147 |    "source": []
148 |   }
149 |  ],
150 |  "metadata": {
151 |   "kernelspec": {
152 |    "display_name": "Python 2",
153 |    "language": "python",
154 |    "name": "python2"
155 |   },
156 |   "language_info": {
157 |    "codemirror_mode": {
158 |     "name": "ipython",
159 |     "version": 2
160 |    },
161 |    "file_extension": ".py",
162 |    "mimetype": "text/x-python",
163 |    "name": "python",
164 |    "nbconvert_exporter": "python",
165 |    "pygments_lexer": "ipython2",
166 |    "version": "2.7.11"
167 |   }
168 |  },
169 |  "nbformat": 4,
170 |  "nbformat_minor": 0
171 | }
172 | 


--------------------------------------------------------------------------------
/MNIST_LSTM.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 1,
  6 |    "metadata": {
  7 |     "collapsed": true
  8 |    },
  9 |    "outputs": [],
 10 |    "source": [
 11 |     "import tensorflow as tf\n",
 12 |     "from tensorflow.examples.tutorials.mnist import input_data"
 13 |    ]
 14 |   },
 15 |   {
 16 |    "cell_type": "code",
 17 |    "execution_count": 2,
 18 |    "metadata": {
 19 |     "collapsed": false
 20 |    },
 21 |    "outputs": [
 22 |     {
 23 |      "name": "stdout",
 24 |      "output_type": "stream",
 25 |      "text": [
 26 |       "Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
 27 |       "Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
 28 |       "Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
 29 |       "Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n"
 30 |      ]
 31 |     }
 32 |    ],
 33 |    "source": [
 34 |     "mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)"
 35 |    ]
 36 |   },
 37 |   {
 38 |    "cell_type": "code",
 39 |    "execution_count": 3,
 40 |    "metadata": {
 41 |     "collapsed": false
 42 |    },
 43 |    "outputs": [],
 44 |    "source": [
 45 |     "LSTM_UNITS = 256\n",
 46 |     "BATCH_SIZE = 100\n",
 47 |     "LEARNING_RATE = 1e-3\n",
 48 |     "N_CLASSES = 10\n",
 49 |     "TRAINING_ITER = 1000\n",
 50 |     "\n",
 51 |     "def LSTM(x):\n",
 52 |     "    \n",
 53 |     "    lstm_cell = tf.contrib.rnn.BasicLSTMCell(num_units=LSTM_UNITS)\n",
 54 |     "    \n",
 55 |     "    init_state = lstm_cell.zero_state(batch_size=BATCH_SIZE, dtype=tf.float32)\n",
 56 |     "    \n",
 57 |     "    outputs, final_state = tf.nn.dynamic_rnn(cell=lstm_cell,inputs=x,initial_state=init_state,time_major=False) \n",
 58 |     "    \n",
 59 |     "    return tf.unstack(tf.transpose(outputs, [1, 0, 2]))\n",
 60 |     "\n",
 61 |     "def init_weights(shape):\n",
 62 |     "    return tf.Variable(tf.truncated_normal(shape, stddev=0.1))\n",
 63 |     "\n",
 64 |     "def init_biases(shape):\n",
 65 |     "    return tf.Variable(tf.zeros(shape) + 0.1)\n",
 66 |     "\n",
 67 |     "def add_layer(x, weights, biases, activation_function=None):\n",
 68 |     "    results = tf.matmul(x, weights) + biases\n",
 69 |     "    if activation_function:\n",
 70 |     "        return activation_function(results)\n",
 71 |     "    else:\n",
 72 |     "        return results\n",
 73 |     "\n",
 74 |     "weights = {'output': init_weights([LSTM_UNITS, N_CLASSES])}\n",
 75 |     "biases = {'output': init_biases([N_CLASSES])}\n",
 76 |     "\n",
 77 |     "x = tf.placeholder(tf.float32, [None, 784])\n",
 78 |     "y = tf.placeholder(tf.float32, [None, N_CLASSES])\n",
 79 |     "\n",
 80 |     "x_2d = tf.reshape(x, [-1, 28, 28])\n",
 81 |     "\n",
 82 |     "lstm_outputs = LSTM(x_2d)\n",
 83 |     "\n",
 84 |     "predictions = add_layer(lstm_outputs[-1], weights['output'], biases['output'], None)\n",
 85 |     "\n",
 86 |     "loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=predictions))\n",
 87 |     "\n",
 88 |     "train_step = tf.train.AdamOptimizer(LEARNING_RATE).minimize(loss)\n",
 89 |     "\n",
 90 |     "accuracy = tf.reduce_mean(tf.cast(tf.equal(tf.arg_max(y, 1), tf.arg_max(predictions, 1)), tf.float32))\n",
 91 |     "\n"
 92 |    ]
 93 |   },
 94 |   {
 95 |    "cell_type": "code",
 96 |    "execution_count": 4,
 97 |    "metadata": {
 98 |     "collapsed": false
 99 |    },
100 |    "outputs": [
101 |     {
102 |      "name": "stdout",
103 |      "output_type": "stream",
104 |      "text": [
105 |       "TRAIN_ACC@0: 0.27\n",
106 |       "VALID_ACC@0:"
107 |      ]
108 |     },
109 |     {
110 |      "ename": "InvalidArgumentError",
111 |      "evalue": "ConcatOp : Dimensions of inputs should match: shape[0] = [5000,28] vs. shape[1] = [100,256]\n\t [[Node: rnn/while/basic_lstm_cell/basic_lstm_cell/concat = ConcatV2[N=2, T=DT_FLOAT, Tidx=DT_INT32, _device=\"/job:localhost/replica:0/task:0/cpu:0\"](rnn/while/TensorArrayReadV3, rnn/while/Identity_3, rnn/while/basic_lstm_cell/basic_lstm_cell/concat/axis)]]\n\nCaused by op u'rnn/while/basic_lstm_cell/basic_lstm_cell/concat', defined at:\n  File \"/Users/miaofan/anaconda/lib/python2.7/runpy.py\", line 162, in _run_module_as_main\n    \"__main__\", fname, loader, pkg_name)\n  File \"/Users/miaofan/anaconda/lib/python2.7/runpy.py\", line 72, in _run_code\n    exec code in run_globals\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/__main__.py\", line 3, in <module>\n    app.launch_new_instance()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/traitlets/config/application.py\", line 658, in launch_instance\n    app.start()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/kernelapp.py\", line 405, in start\n    ioloop.IOLoop.instance().start()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/zmq/eventloop/ioloop.py\", line 162, in start\n    super(ZMQIOLoop, self).start()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tornado/ioloop.py\", line 883, in start\n    handler_func(fd_obj, events)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tornado/stack_context.py\", line 275, in null_wrapper\n    return fn(*args, **kwargs)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py\", line 440, in _handle_events\n    self._handle_recv()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py\", line 472, in _handle_recv\n    self._run_callback(callback, msg)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py\", line 414, in _run_callback\n    callback(*args, **kwargs)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tornado/stack_context.py\", line 275, in null_wrapper\n    return fn(*args, **kwargs)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/kernelbase.py\", line 260, in dispatcher\n    return self.dispatch_shell(stream, msg)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/kernelbase.py\", line 212, in dispatch_shell\n    handler(stream, idents, msg)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/kernelbase.py\", line 370, in execute_request\n    user_expressions, allow_stdin)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/ipkernel.py\", line 175, in do_execute\n    shell.run_cell(code, store_history=store_history, silent=silent)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/IPython/core/interactiveshell.py\", line 2717, in run_cell\n    interactivity=interactivity, compiler=compiler, result=result)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/IPython/core/interactiveshell.py\", line 2821, in run_ast_nodes\n    if self.run_code(code, result):\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/IPython/core/interactiveshell.py\", line 2881, in run_code\n    exec(code_obj, self.user_global_ns, self.user_ns)\n  File \"<ipython-input-3-97985244513f>\", line 38, in <module>\n    lstm_outputs = LSTM(x_2d)\n  File \"<ipython-input-3-97985244513f>\", line 13, in LSTM\n    outputs, final_state = tf.nn.dynamic_rnn(cell=lstm_cell,inputs=x,initial_state=init_state,time_major=False)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py\", line 546, in dynamic_rnn\n    dtype=dtype)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py\", line 713, in _dynamic_rnn_loop\n    swap_memory=swap_memory)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py\", line 2605, in while_loop\n    result = context.BuildLoop(cond, body, loop_vars, shape_invariants)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py\", line 2438, in BuildLoop\n    pred, body, original_loop_vars, loop_vars, shape_invariants)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py\", line 2388, in _BuildLoop\n    body_result = body(*packed_vars_for_body)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py\", line 698, in _time_step\n    (output, new_state) = call_cell()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py\", line 684, in <lambda>\n    call_cell = lambda: cell(input_t, state)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/contrib/rnn/python/ops/core_rnn_cell_impl.py\", line 179, in __call__\n    concat = _linear([inputs, h], 4 * self._num_units, True, scope=scope)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/contrib/rnn/python/ops/core_rnn_cell_impl.py\", line 751, in _linear\n    res = math_ops.matmul(array_ops.concat(args, 1), weights)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/array_ops.py\", line 1034, in concat\n    name=name)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/gen_array_ops.py\", line 519, in _concat_v2\n    name=name)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py\", line 763, in apply_op\n    op_def=op_def)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/ops.py\", line 2327, in create_op\n    original_op=self._default_original_op, op_def=op_def)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/ops.py\", line 1226, in __init__\n    self._traceback = _extract_stack()\n\nInvalidArgumentError (see above for traceback): ConcatOp : Dimensions of inputs should match: shape[0] = [5000,28] vs. shape[1] = [100,256]\n\t [[Node: rnn/while/basic_lstm_cell/basic_lstm_cell/concat = ConcatV2[N=2, T=DT_FLOAT, Tidx=DT_INT32, _device=\"/job:localhost/replica:0/task:0/cpu:0\"](rnn/while/TensorArrayReadV3, rnn/while/Identity_3, rnn/while/basic_lstm_cell/basic_lstm_cell/concat/axis)]]\n",
112 |      "output_type": "error",
113 |      "traceback": [
114 |       "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
115 |       "\u001b[0;31mInvalidArgumentError\u001b[0m                      Traceback (most recent call last)",
116 |       "\u001b[0;32m<ipython-input-4-50c696363f46>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      9\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mi\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0;36m50\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     10\u001b[0m             \u001b[0;32mprint\u001b[0m \u001b[0;34m'TRAIN_ACC@%d:'\u001b[0m \u001b[0;34m%\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msess\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0maccuracy\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mbatch_xs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mbatch_ys\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 11\u001b[0;31m             \u001b[0;32mprint\u001b[0m \u001b[0;34m'VALID_ACC@%d:'\u001b[0m \u001b[0;34m%\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msess\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0maccuracy\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m{\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mmnist\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalidation\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimages\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mmnist\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalidation\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlabels\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
117 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    765\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    766\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 767\u001b[0;31m                          run_metadata_ptr)\n\u001b[0m\u001b[1;32m    768\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    769\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
118 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    963\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    964\u001b[0m       results = self._do_run(handle, final_targets, final_fetches,\n\u001b[0;32m--> 965\u001b[0;31m                              feed_dict_string, options, run_metadata)\n\u001b[0m\u001b[1;32m    966\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    967\u001b[0m       \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
119 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_do_run\u001b[0;34m(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1013\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1014\u001b[0m       return self._do_call(_run_fn, self._session, feed_dict, fetch_list,\n\u001b[0;32m-> 1015\u001b[0;31m                            target_list, options, run_metadata)\n\u001b[0m\u001b[1;32m   1016\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1017\u001b[0m       return self._do_call(_prun_fn, self._session, handle, feed_dict,\n",
120 |       "\u001b[0;32m/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_do_call\u001b[0;34m(self, fn, *args)\u001b[0m\n\u001b[1;32m   1033\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1034\u001b[0m           \u001b[0;32mpass\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1035\u001b[0;31m       \u001b[0;32mraise\u001b[0m \u001b[0mtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode_def\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mop\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmessage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1036\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1037\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0m_extend_graph\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
121 |       "\u001b[0;31mInvalidArgumentError\u001b[0m: ConcatOp : Dimensions of inputs should match: shape[0] = [5000,28] vs. shape[1] = [100,256]\n\t [[Node: rnn/while/basic_lstm_cell/basic_lstm_cell/concat = ConcatV2[N=2, T=DT_FLOAT, Tidx=DT_INT32, _device=\"/job:localhost/replica:0/task:0/cpu:0\"](rnn/while/TensorArrayReadV3, rnn/while/Identity_3, rnn/while/basic_lstm_cell/basic_lstm_cell/concat/axis)]]\n\nCaused by op u'rnn/while/basic_lstm_cell/basic_lstm_cell/concat', defined at:\n  File \"/Users/miaofan/anaconda/lib/python2.7/runpy.py\", line 162, in _run_module_as_main\n    \"__main__\", fname, loader, pkg_name)\n  File \"/Users/miaofan/anaconda/lib/python2.7/runpy.py\", line 72, in _run_code\n    exec code in run_globals\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/__main__.py\", line 3, in <module>\n    app.launch_new_instance()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/traitlets/config/application.py\", line 658, in launch_instance\n    app.start()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/kernelapp.py\", line 405, in start\n    ioloop.IOLoop.instance().start()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/zmq/eventloop/ioloop.py\", line 162, in start\n    super(ZMQIOLoop, self).start()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tornado/ioloop.py\", line 883, in start\n    handler_func(fd_obj, events)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tornado/stack_context.py\", line 275, in null_wrapper\n    return fn(*args, **kwargs)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py\", line 440, in _handle_events\n    self._handle_recv()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py\", line 472, in _handle_recv\n    self._run_callback(callback, msg)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py\", line 414, in _run_callback\n    callback(*args, **kwargs)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tornado/stack_context.py\", line 275, in null_wrapper\n    return fn(*args, **kwargs)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/kernelbase.py\", line 260, in dispatcher\n    return self.dispatch_shell(stream, msg)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/kernelbase.py\", line 212, in dispatch_shell\n    handler(stream, idents, msg)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/kernelbase.py\", line 370, in execute_request\n    user_expressions, allow_stdin)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/ipykernel/ipkernel.py\", line 175, in do_execute\n    shell.run_cell(code, store_history=store_history, silent=silent)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/IPython/core/interactiveshell.py\", line 2717, in run_cell\n    interactivity=interactivity, compiler=compiler, result=result)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/IPython/core/interactiveshell.py\", line 2821, in run_ast_nodes\n    if self.run_code(code, result):\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/IPython/core/interactiveshell.py\", line 2881, in run_code\n    exec(code_obj, self.user_global_ns, self.user_ns)\n  File \"<ipython-input-3-97985244513f>\", line 38, in <module>\n    lstm_outputs = LSTM(x_2d)\n  File \"<ipython-input-3-97985244513f>\", line 13, in LSTM\n    outputs, final_state = tf.nn.dynamic_rnn(cell=lstm_cell,inputs=x,initial_state=init_state,time_major=False)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py\", line 546, in dynamic_rnn\n    dtype=dtype)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py\", line 713, in _dynamic_rnn_loop\n    swap_memory=swap_memory)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py\", line 2605, in while_loop\n    result = context.BuildLoop(cond, body, loop_vars, shape_invariants)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py\", line 2438, in BuildLoop\n    pred, body, original_loop_vars, loop_vars, shape_invariants)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py\", line 2388, in _BuildLoop\n    body_result = body(*packed_vars_for_body)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py\", line 698, in _time_step\n    (output, new_state) = call_cell()\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py\", line 684, in <lambda>\n    call_cell = lambda: cell(input_t, state)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/contrib/rnn/python/ops/core_rnn_cell_impl.py\", line 179, in __call__\n    concat = _linear([inputs, h], 4 * self._num_units, True, scope=scope)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/contrib/rnn/python/ops/core_rnn_cell_impl.py\", line 751, in _linear\n    res = math_ops.matmul(array_ops.concat(args, 1), weights)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/array_ops.py\", line 1034, in concat\n    name=name)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/ops/gen_array_ops.py\", line 519, in _concat_v2\n    name=name)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py\", line 763, in apply_op\n    op_def=op_def)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/ops.py\", line 2327, in create_op\n    original_op=self._default_original_op, op_def=op_def)\n  File \"/Users/miaofan/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/ops.py\", line 1226, in __init__\n    self._traceback = _extract_stack()\n\nInvalidArgumentError (see above for traceback): ConcatOp : Dimensions of inputs should match: shape[0] = [5000,28] vs. shape[1] = [100,256]\n\t [[Node: rnn/while/basic_lstm_cell/basic_lstm_cell/concat = ConcatV2[N=2, T=DT_FLOAT, Tidx=DT_INT32, _device=\"/job:localhost/replica:0/task:0/cpu:0\"](rnn/while/TensorArrayReadV3, rnn/while/Identity_3, rnn/while/basic_lstm_cell/basic_lstm_cell/concat/axis)]]\n"
122 |      ]
123 |     }
124 |    ],
125 |    "source": [
126 |     "init = tf.global_variables_initializer()\n",
127 |     "\n",
128 |     "with tf.Session() as sess:\n",
129 |     "    sess.run(init)\n",
130 |     "    \n",
131 |     "    for i in range(TRAINING_ITER):\n",
132 |     "        batch_xs, batch_ys = mnist.train.next_batch(BATCH_SIZE)\n",
133 |     "        sess.run(train_step, feed_dict= {x: batch_xs, y: batch_ys})\n",
134 |     "        if i % 50 == 0:\n",
135 |     "            print 'TRAIN_ACC@%d:' %i, sess.run(accuracy, feed_dict = {x: batch_xs, y:batch_ys})\n",
136 |     "            \n",
137 |     "            \n",
138 |     "            #print 'VALID_ACC@%d:' %i, sess.run(accuracy, feed_dict = {x: mnist.validation.images, y: mnist.validation.labels})"
139 |    ]
140 |   }
141 |  ],
142 |  "metadata": {
143 |   "kernelspec": {
144 |    "display_name": "Python 2",
145 |    "language": "python",
146 |    "name": "python2"
147 |   },
148 |   "language_info": {
149 |    "codemirror_mode": {
150 |     "name": "ipython",
151 |     "version": 2
152 |    },
153 |    "file_extension": ".py",
154 |    "mimetype": "text/x-python",
155 |    "name": "python",
156 |    "nbconvert_exporter": "python",
157 |    "pygments_lexer": "ipython2",
158 |    "version": "2.7.11"
159 |   }
160 |  },
161 |  "nbformat": 4,
162 |  "nbformat_minor": 0
163 | }
164 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | # Deep_Learning_with_TensorFlow
2 | Deep Learning with TensorFlow implemented in Python mainly practiced with MNIST dataset. 
3 | 


--------------------------------------------------------------------------------