Pushing the docs to dev/ for branch: main, commit f811a03991d0bcd04bc52ee9f659dcd2de1cf4de

Author: dirty-cat-ci

dev/_downloads/00e80c50159fa488ff1ac435ddc159e9/12_subsampling.zip

@@ -237,7 +237,7 @@ def plot_box_results(named_results):
 # case, 30.
 from skrub import StringEncoder
 
-string_encoder = StringEncoder(ngram_range=(3, 4), analyzer="char_wb")
+string_encoder = StringEncoder(ngram_range=(3, 4), analyzer="char_wb", random_state=0)
 
 string_encoder_pipe = make_pipeline(
     TableVectorizer(high_cardinality=string_encoder),
@@ -302,12 +302,13 @@ def plot_performance_tradeoff(results):
         ax.set_title("Prediction performance / training time trade-off")
 
     ax.annotate(
-        "",
-        xy=(1.5, 0.98),
-        xytext=(8.5, 0.90),
-        arrowprops=dict(arrowstyle="->", mutation_scale=15),
+        "Best time / \nperformance trade-off",
+        xy=(0.05, 0.95),
+        xycoords="axes fraction",
+        xytext=(0.2, 0.8),
+        textcoords="axes fraction",
+        arrowprops=dict(arrowstyle="->", lw=1.5, mutation_scale=15),
     )
-    ax.text(5.8, 0.86, "Best time / \nperformance trade-off")
     ax.legend(bbox_to_anchor=(1.02, 0.3))
     plt.show()
 

dev/_downloads/0da1186528765de7400f245b4d95af54/10_expressions_intro.zip

@@ -148,7 +148,7 @@
       },
       "outputs": [],
       "source": [
-        "from skrub import StringEncoder\n\nstring_encoder = StringEncoder(ngram_range=(3, 4), analyzer=\"char_wb\")\n\nstring_encoder_pipe = make_pipeline(\n    TableVectorizer(high_cardinality=string_encoder),\n    HistGradientBoostingClassifier(),\n)\n\nstring_encoder_results = cross_validate(string_encoder_pipe, X, y, scoring=\"roc_auc\")\nresults.append((\"StringEncoder\", string_encoder_results))\n\nplot_box_results(results)"
+        "from skrub import StringEncoder\n\nstring_encoder = StringEncoder(ngram_range=(3, 4), analyzer=\"char_wb\", random_state=0)\n\nstring_encoder_pipe = make_pipeline(\n    TableVectorizer(high_cardinality=string_encoder),\n    HistGradientBoostingClassifier(),\n)\n\nstring_encoder_results = cross_validate(string_encoder_pipe, X, y, scoring=\"roc_auc\")\nresults.append((\"StringEncoder\", string_encoder_results))\n\nplot_box_results(results)"
       ]
     },
     {
@@ -166,7 +166,7 @@
       },
       "outputs": [],
       "source": [
-        "def plot_performance_tradeoff(results):\n    fig, ax = plt.subplots(figsize=(5, 4), dpi=200)\n    markers = [\"s\", \"o\", \"^\", \"x\"]\n    for idx, (name, result) in enumerate(results):\n        ax.scatter(\n            result[\"fit_time\"],\n            result[\"test_score\"],\n            label=name,\n            marker=markers[idx],\n        )\n        mean_fit_time = np.mean(result[\"fit_time\"])\n        mean_score = np.mean(result[\"test_score\"])\n        ax.scatter(\n            mean_fit_time,\n            mean_score,\n            color=\"k\",\n            marker=markers[idx],\n        )\n        std_fit_time = np.std(result[\"fit_time\"])\n        std_score = np.std(result[\"test_score\"])\n        ax.errorbar(\n            x=mean_fit_time,\n            y=mean_score,\n            yerr=std_score,\n            fmt=\"none\",\n            c=\"k\",\n            capsize=2,\n        )\n        ax.errorbar(\n            x=mean_fit_time,\n            y=mean_score,\n            xerr=std_fit_time,\n            fmt=\"none\",\n            c=\"k\",\n            capsize=2,\n        )\n        ax.set_xscale(\"log\")\n\n        ax.set_xlabel(\"Time to fit (seconds)\")\n        ax.set_ylabel(\"ROC AUC\")\n        ax.set_title(\"Prediction performance / training time trade-off\")\n\n    ax.annotate(\n        \"\",\n        xy=(1.5, 0.98),\n        xytext=(8.5, 0.90),\n        arrowprops=dict(arrowstyle=\"->\", mutation_scale=15),\n    )\n    ax.text(5.8, 0.86, \"Best time / \\nperformance trade-off\")\n    ax.legend(bbox_to_anchor=(1.02, 0.3))\n    plt.show()\n\n\nplot_performance_tradeoff(results)"
+        "def plot_performance_tradeoff(results):\n    fig, ax = plt.subplots(figsize=(5, 4), dpi=200)\n    markers = [\"s\", \"o\", \"^\", \"x\"]\n    for idx, (name, result) in enumerate(results):\n        ax.scatter(\n            result[\"fit_time\"],\n            result[\"test_score\"],\n            label=name,\n            marker=markers[idx],\n        )\n        mean_fit_time = np.mean(result[\"fit_time\"])\n        mean_score = np.mean(result[\"test_score\"])\n        ax.scatter(\n            mean_fit_time,\n            mean_score,\n            color=\"k\",\n            marker=markers[idx],\n        )\n        std_fit_time = np.std(result[\"fit_time\"])\n        std_score = np.std(result[\"test_score\"])\n        ax.errorbar(\n            x=mean_fit_time,\n            y=mean_score,\n            yerr=std_score,\n            fmt=\"none\",\n            c=\"k\",\n            capsize=2,\n        )\n        ax.errorbar(\n            x=mean_fit_time,\n            y=mean_score,\n            xerr=std_fit_time,\n            fmt=\"none\",\n            c=\"k\",\n            capsize=2,\n        )\n        ax.set_xscale(\"log\")\n\n        ax.set_xlabel(\"Time to fit (seconds)\")\n        ax.set_ylabel(\"ROC AUC\")\n        ax.set_title(\"Prediction performance / training time trade-off\")\n\n    ax.annotate(\n        \"Best time / \\nperformance trade-off\",\n        xy=(0.05, 0.95),\n        xycoords=\"axes fraction\",\n        xytext=(0.2, 0.8),\n        textcoords=\"axes fraction\",\n        arrowprops=dict(arrowstyle=\"->\", lw=1.5, mutation_scale=15),\n    )\n    ax.legend(bbox_to_anchor=(1.02, 0.3))\n    plt.show()\n\n\nplot_performance_tradeoff(results)"
       ]
     },
     {