Updating Midterm 2 solutions

Midterms/2016 Midterm 2/2016_Midterm_2_solutions.ipynb

@@ -1658,128 +1658,6 @@
    "source": [
     "Yay! We beat the benchmarks on the observed data and did pretty good on test data!"
    ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "# Autograding function"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 33,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [],
-   "source": [
-    "accuracy = lambda y_true, y_pred: (((y_true - y_pred) == 0).sum() * 1.) / len(y_true)\n",
-    "\n",
-    "# function to check - did they beat our benchmarks?\n",
-    "# You can either pass it the flu_predict function the student wrote or the\n",
-    "# file name of the y-labels.\n",
-    "#\n",
-    "# predict: function of type (array -> array)\n",
-    "# pred_y_file_name: file name of where their predicted y-labels live\n",
-    "# data_preprocessing: if false x-train data will be fed to flu_predict with no processing if true data will be encoded with 0 filled in for NaN\n",
-    "# cost: if true computes the expect cost\n",
-    "# \n",
-    "# return some string indicating result of comparison with benchmark\n",
-    "\n",
-    "def beat_benchmark(flu_predict=None, pred_y_file_name=None, data_preprocessing=False):\n",
-    "    \n",
-    "    acc_bm_0 = .69\n",
-    "    acc_bm_1 = .60\n",
-    "    \n",
-    "    acc_rm_0 = .5\n",
-    "    acc_rm_1 = .5\n",
-    "    \n",
-    "    df_test = pd.read_csv('data/flu_test.csv')\n",
-    "    df_test = df_test[~np.isnan(df_test['flu'])]\n",
-    "    df_test['flutype'] = df_test['flutype'] - 1\n",
-    "    \n",
-    "    y_true = df_test.values[:, -2]\n",
-    "    \n",
-    "    if flu_predict is not None:\n",
-    "        if data_preprocessing:                       \n",
-    "            encode = preprocessing.LabelEncoder()\n",
-    "            for column in df_test.columns:\n",
-    "                if df_test[column].dtype == np.object:\n",
-    "                    df_test.loc[:, column] = encode.fit_transform(df_test[column])\n",
-    "\n",
-    "            df_test = df_test.fillna(0)\n",
-    "            \n",
-    "        x = df_test.values[:, :-2]\n",
-    "        \n",
-    "        y_pred = flu_predict(x)\n",
-    "        \n",
-    "    elif pred_y_file_name is not None:\n",
-    "        df_y_pred = pd.read_csv(pred_y_file_name)\n",
-    "        y_pred = df_y_pred.values[:, -1]\n",
-    "        \n",
-    "    else:\n",
-    "        return 'params ill-specified'\n",
-    "    \n",
-    "    acc_0 = accuracy(y_true[y_true == 0], y_pred[y_true == 0])\n",
-    "    acc_1 = accuracy(y_true[y_true == 1], y_pred[y_true == 1])\n",
-    "    \n",
-    "    if acc_0 > acc_rm_0 and acc_1 > acc_rm_1:\n",
-    "        print ('accuracies: {}, {}'.format(acc_0, acc_1))\n",
-    "        return \"accuracy: beats all benchmarks :)\"\n",
-    "    elif acc_0 > acc_bm_0 and acc_1 > acc_bm_1:\n",
-    "        print ('accuracies: {}, {}'.format(acc_0, acc_1))\n",
-    "        return \"accuracy: beats only baseline models :/\"\n",
-    "    else:\n",
-    "        print ('accuracies: {}, {}'.format(acc_0, acc_1))\n",
-    "        return \"accuracy: beats no benchmarks :(\""
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 34,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "accuracies: 1.0, 0.0\n",
-      "accuracy: beats no benchmarks :(\n"
-     ]
-    }
-   ],
-   "source": [
-    "# Example 0: using the autograding function\n",
-    "\n",
-    "def flu_predict(x):\n",
-    "    return np.array([0] * len(x))\n",
-    "\n",
-    "print(beat_benchmark(flu_predict, data_preprocessing=False))"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 35,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "accuracies: 0.7191558441558441, 0.6112956810631229\n",
-      "accuracy: beats all benchmarks :)\n"
-     ]
-    }
-   ],
-   "source": [
-    "# Example 1: using the autograding function\n",
-    "print(beat_benchmark(pred_y_file_name='data/Example_1.csv'))"
-   ]
   }
  ],
  "metadata": {