Pushing the docs to dev/ for branch: master, commit 10a05a368542a7c1afeadf29b666a51166b9b0b3

Author: skoptci

dev/_downloads/3409d9766d352cc9f9b169d4a799a87a/auto_examples_python.zip

@@ -26,7 +26,7 @@
       },
       "outputs": [],
       "source": [
-        "print(__doc__)\nimport numpy as np"
+        "print(__doc__)\nimport numpy as np\nnp.random.seed(123)\nimport matplotlib.pyplot as plt"
       ]
     },
     {
@@ -62,7 +62,43 @@
       },
       "outputs": [],
       "source": [
-        "from skopt import BayesSearchCV\nfrom skopt.space import Real, Categorical, Integer\n\nfrom sklearn.datasets import load_digits\nfrom sklearn.svm import LinearSVC, SVC\nfrom sklearn.pipeline import Pipeline\nfrom sklearn.model_selection import train_test_split\n\nX, y = load_digits(10, True)\nX_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)\n\n# pipeline class is used as estimator to enable\n# search over different model types\npipe = Pipeline([\n    ('model', SVC())\n])\n\n# single categorical value of 'model' parameter is\n# sets the model class\n# We will get ConvergenceWarnings because the problem is not well-conditioned.\n# But that's fine, this is just an example.\nlinsvc_search = {\n    'model': [LinearSVC(max_iter=1000)],\n    'model__C': (1e-6, 1e+6, 'log-uniform'),\n}\n\n# explicit dimension classes can be specified like this\nsvc_search = {\n    'model': Categorical([SVC()]),\n    'model__C': Real(1e-6, 1e+6, prior='log-uniform'),\n    'model__gamma': Real(1e-6, 1e+1, prior='log-uniform'),\n    'model__degree': Integer(1,8),\n    'model__kernel': Categorical(['linear', 'poly', 'rbf']),\n}\n\nopt = BayesSearchCV(\n    pipe,\n    [(svc_search, 20), (linsvc_search, 16)], # (parameter space, # of evaluations)\n    cv=3\n)\n\nopt.fit(X_train, y_train)\n\nprint(\"val. score: %s\" % opt.best_score_)\nprint(\"test score: %s\" % opt.score(X_test, y_test))"
+        "from skopt import BayesSearchCV\nfrom skopt.space import Real, Categorical, Integer\nfrom skopt.plots import plot_objective, plot_histogram\n\nfrom sklearn.datasets import load_digits\nfrom sklearn.svm import LinearSVC, SVC\nfrom sklearn.pipeline import Pipeline\nfrom sklearn.model_selection import train_test_split\n\nX, y = load_digits(10, True)\nX_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)\n\n# pipeline class is used as estimator to enable\n# search over different model types\npipe = Pipeline([\n    ('model', SVC())\n])\n\n# single categorical value of 'model' parameter is\n# sets the model class\n# We will get ConvergenceWarnings because the problem is not well-conditioned.\n# But that's fine, this is just an example.\nlinsvc_search = {\n    'model': [LinearSVC(max_iter=1000)],\n    'model__C': (1e-6, 1e+6, 'log-uniform'),\n}\n\n# explicit dimension classes can be specified like this\nsvc_search = {\n    'model': Categorical([SVC()]),\n    'model__C': Real(1e-6, 1e+6, prior='log-uniform'),\n    'model__gamma': Real(1e-6, 1e+1, prior='log-uniform'),\n    'model__degree': Integer(1,8),\n    'model__kernel': Categorical(['linear', 'poly', 'rbf']),\n}\n\nopt = BayesSearchCV(\n    pipe,\n    # (parameter space, # of evaluations)\n    [(svc_search, 40), (linsvc_search, 16)],\n    cv=3\n)\n\nopt.fit(X_train, y_train)\n\nprint(\"val. score: %s\" % opt.best_score_)\nprint(\"test score: %s\" % opt.score(X_test, y_test))\nprint(\"best params: %s\" % str(opt.best_params_))"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "Partial Dependence plot of the objective function for SVC\n\n\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "_ = plot_objective(opt.optimizer_results_[0],\n                   dimensions=[\"C\", \"degree\", \"gamma\", \"kernel\"],\n                   n_minimum_search=int(1e8))\nplt.show()"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "Plot of the histogram for LinearSVC\n\n\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "collapsed": false
+      },
+      "outputs": [],
+      "source": [
+        "_ = plot_histogram(opt.optimizer_results_[1], 1)\nplt.show()"
       ]
     },
     {
@@ -80,7 +116,7 @@
       },
       "outputs": [],
       "source": [
-        "from skopt import BayesSearchCV\n\nfrom sklearn.datasets import load_iris\nfrom sklearn.svm import SVC\n\nX, y = load_iris(True)\n\nsearchcv = BayesSearchCV(\n    SVC(gamma='scale'),\n    search_spaces={'C': (0.01, 100.0, 'log-uniform')},\n    n_iter=10,\n    cv=3\n)\n\n\n# callback handler\ndef on_step(optim_result):\n    score = searchcv.best_score_\n    print(\"best score: %s\" % score)\n    if score >= 0.98:\n        print('Interrupting!')\n        return True\n\n\nsearchcv.fit(X, y, callback=on_step)"
+        "from skopt import BayesSearchCV\n\nfrom sklearn.datasets import load_iris\nfrom sklearn.svm import SVC\n\nX, y = load_iris(True)\n\nsearchcv = BayesSearchCV(\n    SVC(gamma='scale'),\n    search_spaces={'C': (0.01, 100.0, 'log-uniform')},\n    n_iter=10,\n    cv=3\n)\n\n# callback handler\ndef on_step(optim_result):\n    score = searchcv.best_score_\n    print(\"best score: %s\" % score)\n    if score >= 0.98:\n        print('Interrupting!')\n        return True\n\n\nsearchcv.fit(X, y, callback=on_step)"
       ]
     },
     {

dev/_downloads/6bf99924122eedf93b886bae30f1182b/sklearn-gridsearchcv-replacement.ipynb

@@ -35,6 +35,8 @@
 """
 print(__doc__)
 import numpy as np
+np.random.seed(123)
+import matplotlib.pyplot as plt
 
 #############################################################################
 # Minimal example
@@ -79,6 +81,7 @@
 
 from skopt import BayesSearchCV
 from skopt.space import Real, Categorical, Integer
+from skopt.plots import plot_objective, plot_histogram
 
 from sklearn.datasets import load_digits
 from sklearn.svm import LinearSVC, SVC
@@ -114,14 +117,30 @@
 
 opt = BayesSearchCV(
     pipe,
-    [(svc_search, 20), (linsvc_search, 16)], # (parameter space, # of evaluations)
+    # (parameter space, # of evaluations)
+    [(svc_search, 40), (linsvc_search, 16)],
     cv=3
 )
 
 opt.fit(X_train, y_train)
 
 print("val. score: %s" % opt.best_score_)
 print("test score: %s" % opt.score(X_test, y_test))
+print("best params: %s" % str(opt.best_params_))
+
+#############################################################################
+# Partial Dependence plot of the objective function for SVC
+#
+_ = plot_objective(opt.optimizer_results_[0],
+                   dimensions=["C", "degree", "gamma", "kernel"],
+                   n_minimum_search=int(1e8))
+plt.show()
+
+#############################################################################
+# Plot of the histogram for LinearSVC
+#
+_ = plot_histogram(opt.optimizer_results_[1], 1)
+plt.show()
 
 #############################################################################
 # Progress monitoring and control using `callback` argument of `fit` method
@@ -153,7 +172,6 @@
     cv=3
 )
 
-
 # callback handler
 def on_step(optim_result):
     score = searchcv.best_score_

dev/_downloads/d34667f097c619f8afda4bc936e7af21/auto_examples_jupyter.zip

@@ -295,8 +295,10 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
 
     <span class="k">if</span> <span class="n">ax</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
         <span class="n">ax</span> <span class="o">=</span> <span class="n">plt</span><span class="o">.</span><span class="n">gca</span><span class="p">()</span>
-    <span class="k">assert</span> <span class="n">res</span><span class="o">.</span><span class="n">space</span><span class="o">.</span><span class="n">n_dims</span> <span class="o">==</span> <span class="mi">1</span><span class="p">,</span> <span class="s2">&quot;Space dimension must be 1&quot;</span>
-    <span class="n">x</span><span class="p">,</span> <span class="n">x_model</span> <span class="o">=</span> <span class="n">_evenly_sample</span><span class="p">(</span><span class="n">res</span><span class="o">.</span><span class="n">space</span><span class="o">.</span><span class="n">dimensions</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">n_points</span><span class="p">)</span>
+    <span class="n">n_dims</span> <span class="o">=</span> <span class="n">res</span><span class="o">.</span><span class="n">space</span><span class="o">.</span><span class="n">n_dims</span>
+    <span class="k">assert</span> <span class="n">n_dims</span> <span class="o">==</span> <span class="mi">1</span><span class="p">,</span> <span class="s2">&quot;Space dimension must be 1&quot;</span>
+    <span class="n">dimension</span> <span class="o">=</span> <span class="n">res</span><span class="o">.</span><span class="n">space</span><span class="o">.</span><span class="n">dimensions</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
+    <span class="n">x</span><span class="p">,</span> <span class="n">x_model</span> <span class="o">=</span> <span class="n">_evenly_sample</span><span class="p">(</span><span class="n">dimension</span><span class="p">,</span> <span class="n">n_points</span><span class="p">)</span>
     <span class="n">x</span> <span class="o">=</span> <span class="n">x</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
     <span class="n">x_model</span> <span class="o">=</span> <span class="n">x_model</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
     <span class="k">if</span> <span class="n">res</span><span class="o">.</span><span class="n">specs</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span> <span class="ow">and</span> <span class="s2">&quot;args&quot;</span> <span class="ow">in</span> <span class="n">res</span><span class="o">.</span><span class="n">specs</span><span class="p">:</span>
@@ -736,7 +738,8 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
 <span class="sd">    plot_dims : list of str and int, default=None</span>
 <span class="sd">        List of dimension names or dimension indices from the</span>
 <span class="sd">        search-space dimensions to be included in the plot.</span>
-<span class="sd">        If `None` then use all dimensions from the search-space.</span>
+<span class="sd">        If `None` then use all dimensions except constant ones</span>
+<span class="sd">        from the search-space.</span>
 
 <span class="sd">    sample_source : str or list of floats, default=&#39;random&#39;</span>
 <span class="sd">        Defines to samples generation to use for averaging the model function</span>
@@ -750,31 +753,26 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
 
 <span class="sd">        Valid strings:</span>
 
-<span class="sd">            - &#39;random&#39; - `n_samples` random samples will used</span>
-
-<span class="sd">            - &#39;result&#39; - Use only the best observed parameters</span>
-
-<span class="sd">            - &#39;expected_minimum&#39; - Parameters that gives the best</span>
-<span class="sd">                  minimum Calculated using scipy&#39;s minimize method.</span>
-<span class="sd">                  This method currently does not work with categorical values.</span>
-
-<span class="sd">            - &#39;expected_minimum_random&#39; - Parameters that gives the</span>
-<span class="sd">                  best minimum when using naive random sampling.</span>
-<span class="sd">                  Works with categorical values.</span>
+<span class="sd">        - &#39;random&#39; - `n_samples` random samples will used</span>
+<span class="sd">        - &#39;result&#39; - Use only the best observed parameters</span>
+<span class="sd">        - &#39;expected_minimum&#39; - Parameters that gives the best</span>
+<span class="sd">          minimum Calculated using scipy&#39;s minimize method.</span>
+<span class="sd">          This method currently does not work with categorical values.</span>
+<span class="sd">        - &#39;expected_minimum_random&#39; - Parameters that gives the</span>
+<span class="sd">          best minimum when using naive random sampling.</span>
+<span class="sd">          Works with categorical values.</span>
 
 <span class="sd">    minimum : str or list of floats, default = &#39;result&#39;</span>
 <span class="sd">        Defines the values for the red points in the plots.</span>
 <span class="sd">        Valid strings:</span>
 
-<span class="sd">            - &#39;result&#39; - Use best observed parameters</span>
-
-<span class="sd">            - &#39;expected_minimum&#39; - Parameters that gives the best</span>
-<span class="sd">                  minimum Calculated using scipy&#39;s minimize method.</span>
-<span class="sd">                  This method currently does not work with categorical values.</span>
-
-<span class="sd">            - &#39;expected_minimum_random&#39; - Parameters that gives the</span>
-<span class="sd">                  best minimum when using naive random sampling.</span>
-<span class="sd">                  Works with categorical values</span>
+<span class="sd">        - &#39;result&#39; - Use best observed parameters</span>
+<span class="sd">        - &#39;expected_minimum&#39; - Parameters that gives the best</span>
+<span class="sd">          minimum Calculated using scipy&#39;s minimize method.</span>
+<span class="sd">          This method currently does not work with categorical values.</span>
+<span class="sd">        - &#39;expected_minimum_random&#39; - Parameters that gives the</span>
+<span class="sd">          best minimum when using naive random sampling.</span>
+<span class="sd">          Works with categorical values</span>
 
 <span class="sd">    n_minimum_search : int, default = None</span>
 <span class="sd">        Determines how many points should be evaluated</span>
@@ -787,6 +785,7 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
 <span class="sd">    -------</span>
 <span class="sd">    ax : `Matplotlib.Axes`</span>
 <span class="sd">        A 2-d matrix of Axes-objects with the sub-plots.</span>
+
 <span class="sd">    &quot;&quot;&quot;</span>
     <span class="c1"># Here we define the values for which to plot the red dot (2d plot) and</span>
     <span class="c1"># the red dotted line (1d plot).</span>
@@ -799,6 +798,8 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
         <span class="c1"># Get all dimensions.</span>
         <span class="n">plot_dims</span> <span class="o">=</span> <span class="p">[]</span>
         <span class="k">for</span> <span class="n">row</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">space</span><span class="o">.</span><span class="n">n_dims</span><span class="p">):</span>
+            <span class="k">if</span> <span class="n">space</span><span class="o">.</span><span class="n">dimensions</span><span class="p">[</span><span class="n">row</span><span class="p">]</span><span class="o">.</span><span class="n">is_constant</span><span class="p">:</span>
+                <span class="k">continue</span>
             <span class="n">plot_dims</span><span class="o">.</span><span class="n">append</span><span class="p">((</span><span class="n">row</span><span class="p">,</span> <span class="n">space</span><span class="o">.</span><span class="n">dimensions</span><span class="p">[</span><span class="n">row</span><span class="p">]))</span>
     <span class="k">else</span><span class="p">:</span>
         <span class="n">plot_dims</span> <span class="o">=</span> <span class="n">space</span><span class="p">[</span><span class="n">plot_dims</span><span class="p">]</span>
@@ -898,12 +899,14 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
 <span class="sd">    plot_dims : list of str and int, default=None</span>
 <span class="sd">        List of dimension names or dimension indices from the</span>
 <span class="sd">        search-space dimensions to be included in the plot.</span>
-<span class="sd">        If `None` then use all dimensions from the search-space.</span>
+<span class="sd">        If `None` then use all dimensions except constant ones</span>
+<span class="sd">        from the search-space.</span>
 
 <span class="sd">    Returns</span>
 <span class="sd">    -------</span>
 <span class="sd">    ax : `Matplotlib.Axes`</span>
 <span class="sd">        A 2-d matrix of Axes-objects with the sub-plots.</span>
+
 <span class="sd">    &quot;&quot;&quot;</span>
     <span class="n">space</span> <span class="o">=</span> <span class="n">result</span><span class="o">.</span><span class="n">space</span>
     <span class="c1"># Convert categoricals to integers, so we can ensure consistent ordering.</span>
@@ -917,6 +920,8 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
         <span class="c1"># Get all dimensions.</span>
         <span class="n">plot_dims</span> <span class="o">=</span> <span class="p">[]</span>
         <span class="k">for</span> <span class="n">row</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">space</span><span class="o">.</span><span class="n">n_dims</span><span class="p">):</span>
+            <span class="k">if</span> <span class="n">space</span><span class="o">.</span><span class="n">dimensions</span><span class="p">[</span><span class="n">row</span><span class="p">]</span><span class="o">.</span><span class="n">is_constant</span><span class="p">:</span>
+                <span class="k">continue</span>
             <span class="n">plot_dims</span><span class="o">.</span><span class="n">append</span><span class="p">((</span><span class="n">row</span><span class="p">,</span> <span class="n">space</span><span class="o">.</span><span class="n">dimensions</span><span class="p">[</span><span class="n">row</span><span class="p">]))</span>
     <span class="k">else</span><span class="p">:</span>
         <span class="n">plot_dims</span> <span class="o">=</span> <span class="n">space</span><span class="p">[</span><span class="n">plot_dims</span><span class="p">]</span>
@@ -979,6 +984,7 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
 <span class="sd">    -------</span>
 <span class="sd">    ylim_diagonal : tuple(int)</span>
 <span class="sd">        The common min and max ylim for the diagonal plots.</span>
+
 <span class="sd">    &quot;&quot;&quot;</span>
 
     <span class="c1"># Number of search-space dimensions used in this plot.</span>
@@ -1040,6 +1046,7 @@ <h1>Source code for skopt.plots</h1><div class="highlight"><pre>
 <span class="sd">    yi : np.array</span>
 <span class="sd">        The average value of the modelled objective function at</span>
 <span class="sd">        each point `xi`.</span>
+
 <span class="sd">    &quot;&quot;&quot;</span>
     <span class="c1"># The idea is to step through one dimension, evaluating the model with</span>
     <span class="c1"># that dimension fixed and averaging either over random values or over</span>