bugfix in sklearn svr notebook, fixes #48

claesenm · Jul 15, 2015 · 66f5e15 · 66f5e15
1 parent 4662b9c
commit 66f5e15
Show file tree

Hide file tree

Showing 2 changed files with 35 additions and 38 deletions.
diff --git a/docs/notebooks/notebooks/sklearn-svr.rst b/docs/notebooks/notebooks/sklearn-svr.rst
@@ -18,8 +18,6 @@ for support vector regression, more specifically:
     import optunity
     import optunity.metrics
     import sklearn.svm
-    import matplotlib.pylab as plt
-    import time
 We start by creating the data set. We use sklearn's diabetes data.
 
 .. code:: python
@@ -70,7 +68,7 @@ estimate of MSE for an SVR with default hyperparameters.
 
 .. parsed-literal::
 
-    6056.311654865084
+    6190.481497665955
 
 
 
@@ -112,16 +110,16 @@ cross-validation.
 
 .. parsed-literal::
 
-    optimal hyperparameters: {'C': 8103.600260416671, 'coef0': 0.5978059895833331, 'degree': 4.797399399165802}
-    optimal hyperparameters: {'C': 19755.083174237032, 'coef0': 0.4450483090495349, 'degree': 4.6829006570233025}
-    optimal hyperparameters: {'C': 8680.403645833334, 'coef0': 0.47259114583333317, 'degree': 3.0486328125000006}
+    optimal hyperparameters: {'C': 12078.673881034498, 'coef0': 0.5011052085197018, 'degree': 4.60890281463418}
+    optimal hyperparameters: {'C': 14391.165364583334, 'coef0': 0.17313151041666666, 'degree': 2.35826171875}
+    optimal hyperparameters: {'C': 11713.456382191061, 'coef0': 0.49836486667796476, 'degree': 4.616077904035152}
 
 
 
 
 .. parsed-literal::
 
-    3122.5387612208156
+    3047.035965991627
 
 
 
@@ -162,16 +160,16 @@ twice-iterated 5-fold cross-validation.
 
 .. parsed-literal::
 
-    optimal hyperparameters: {'C': 81.98731244487078, 'gamma': 4.895475395394894}
-    optimal hyperparameters: {'C': 52.067311507783465, 'gamma': 6.5244315086089815}
-    optimal hyperparameters: {'C': 26.993648162210402, 'gamma': 25.997204804202134}
+    optimal hyperparameters: {'C': 21.654003906250026, 'gamma': 16.536188056152554}
+    optimal hyperparameters: {'C': 80.89867187499999, 'gamma': 3.2346692538501784}
+    optimal hyperparameters: {'C': 19.35431640625002, 'gamma': 22.083848774716085}
 
 
 
 
 .. parsed-literal::
 
-    2982.6835697931674
+    2990.8572696483493
 
 
 
@@ -244,17 +242,18 @@ hyperparameterization.
 
 .. parsed-literal::
 
-    optimal hyperparameters: {'kernel': 'rbf', 'C': 36.396896096116805, 'gamma': 10.015489219932745}
-    optimal hyperparameters: {'kernel': 'rbf', 'C': 25.6858037455125, 'gamma': 17.48771066406458}
-    optimal hyperparameters: {'kernel': 'rbf', 'C': 93.31245007796964, 'gamma': 7.082799132257288}
+    optimal hyperparameters: {'kernel': 'rbf', 'C': 33.70116043112164, 'gamma': 16.32317353448437}
+    optimal hyperparameters: {'kernel': 'rbf', 'C': 58.11404170763237, 'gamma': 26.45349823062099}
+    optimal hyperparameters: {'kernel': 'poly', 'C': 14964.421875843143, 'coef0': 0.5127175861493205, 'degree': 4.045210787998622}
 
 
 
 
 .. parsed-literal::
 
-    3001.3258583571483
+    3107.625560844859
 
 
 
-Looks like an RBF kernel was indeed the best choice!
+It looks like the RBF and polynomial kernel are competitive for this
+problem.
diff --git a/notebooks/sklearn-svr.ipynb b/notebooks/sklearn-svr.ipynb
@@ -1,7 +1,7 @@
 {
  "metadata": {
   "name": "",
-  "signature": "sha256:ffffcd755c9a4dc22df222af2da80e7fa04f94ea53fc9391ecb8f742d370b57d"
+  "signature": "sha256:622e6696890c825b50112f76f68f1961aab9e55f39eb0e5b47e7352d51d29fec"
  },
  "nbformat": 3,
  "nbformat_minor": 0,
@@ -37,9 +37,7 @@
       "import itertools\n",
       "import optunity\n",
       "import optunity.metrics\n",
-      "import sklearn.svm\n",
-      "import matplotlib.pylab as plt\n",
-      "import time"
+      "import sklearn.svm"
      ],
      "language": "python",
      "metadata": {},
@@ -131,7 +129,7 @@
        "output_type": "pyout",
        "prompt_number": 4,
        "text": [
-        "6056.311654865084"
+        "6190.481497665955"
        ]
       }
      ],
@@ -194,16 +192,16 @@
        "output_type": "stream",
        "stream": "stdout",
        "text": [
-        "optimal hyperparameters: {'C': 8103.600260416671, 'coef0': 0.5978059895833331, 'degree': 4.797399399165802}\n",
-        "optimal hyperparameters: {'C': 19755.083174237032, 'coef0': 0.4450483090495349, 'degree': 4.6829006570233025}"
+        "optimal hyperparameters: {'C': 12078.673881034498, 'coef0': 0.5011052085197018, 'degree': 4.60890281463418}\n",
+        "optimal hyperparameters: {'C': 14391.165364583334, 'coef0': 0.17313151041666666, 'degree': 2.35826171875}"
        ]
       },
       {
        "output_type": "stream",
        "stream": "stdout",
        "text": [
         "\n",
-        "optimal hyperparameters: {'C': 8680.403645833334, 'coef0': 0.47259114583333317, 'degree': 3.0486328125000006}"
+        "optimal hyperparameters: {'C': 11713.456382191061, 'coef0': 0.49836486667796476, 'degree': 4.616077904035152}"
        ]
       },
       {
@@ -218,7 +216,7 @@
        "output_type": "pyout",
        "prompt_number": 6,
        "text": [
-        "3122.5387612208156"
+        "3047.035965991627"
        ]
       }
      ],
@@ -281,16 +279,16 @@
        "output_type": "stream",
        "stream": "stdout",
        "text": [
-        "optimal hyperparameters: {'C': 81.98731244487078, 'gamma': 4.895475395394894}\n",
-        "optimal hyperparameters: {'C': 52.067311507783465, 'gamma': 6.5244315086089815}"
+        "optimal hyperparameters: {'C': 21.654003906250026, 'gamma': 16.536188056152554}\n",
+        "optimal hyperparameters: {'C': 80.89867187499999, 'gamma': 3.2346692538501784}"
        ]
       },
       {
        "output_type": "stream",
        "stream": "stdout",
        "text": [
         "\n",
-        "optimal hyperparameters: {'C': 26.993648162210402, 'gamma': 25.997204804202134}"
+        "optimal hyperparameters: {'C': 19.35431640625002, 'gamma': 22.083848774716085}"
        ]
       },
       {
@@ -305,7 +303,7 @@
        "output_type": "pyout",
        "prompt_number": 8,
        "text": [
-        "2982.6835697931674"
+        "2990.8572696483493"
        ]
       }
      ],
@@ -348,7 +346,7 @@
      "language": "python",
      "metadata": {},
      "outputs": [],
-     "prompt_number": 21
+     "prompt_number": 9
     },
     {
      "cell_type": "markdown",
@@ -398,7 +396,7 @@
      "language": "python",
      "metadata": {},
      "outputs": [],
-     "prompt_number": 26
+     "prompt_number": 10
     },
     {
      "cell_type": "markdown",
@@ -420,16 +418,16 @@
        "output_type": "stream",
        "stream": "stdout",
        "text": [
-        "optimal hyperparameters: {'kernel': 'rbf', 'C': 36.396896096116805, 'gamma': 10.015489219932745}\n",
-        "optimal hyperparameters: {'kernel': 'rbf', 'C': 25.6858037455125, 'gamma': 17.48771066406458}"
+        "optimal hyperparameters: {'kernel': 'rbf', 'C': 33.70116043112164, 'gamma': 16.32317353448437}\n",
+        "optimal hyperparameters: {'kernel': 'rbf', 'C': 58.11404170763237, 'gamma': 26.45349823062099}"
        ]
       },
       {
        "output_type": "stream",
        "stream": "stdout",
        "text": [
         "\n",
-        "optimal hyperparameters: {'kernel': 'rbf', 'C': 93.31245007796964, 'gamma': 7.082799132257288}"
+        "optimal hyperparameters: {'kernel': 'poly', 'C': 14964.421875843143, 'coef0': 0.5127175861493205, 'degree': 4.045210787998622}"
        ]
       },
       {
@@ -442,19 +440,19 @@
       {
        "metadata": {},
        "output_type": "pyout",
-       "prompt_number": 27,
+       "prompt_number": 13,
        "text": [
-        "3001.3258583571483"
+        "3107.625560844859"
        ]
       }
      ],
-     "prompt_number": 27
+     "prompt_number": 13
     },
     {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
-      "Looks like an RBF kernel was indeed the best choice!"
+      "It looks like the RBF and polynomial kernel are competitive for this problem."
      ]
     }
    ],