Benchmark with full data: Figure 1 notebook

docs/tutorials/functions/unc_forest_tutorials_functions.py

@@ -156,5 +156,5 @@ def plot_fig1(algos, num_plotted_trials, X_eval):
     axes[3].set_ylabel(r"Var($\hat P(Y = 1|X = x)$)")
 
     fig.tight_layout()
-    plt.savefig("fig1.pdf")
+    # plt.savefig("fig1.pdf")
     plt.show()

proglearn/forest.py

@@ -245,17 +245,28 @@ class UncertaintyForest:
     max_depth : int, default=30
         The maximum depth of a tree in the UncertaintyForest
 
+    tree_construction_proportion : float, default = 0.67
+        The proportions of the input data set aside to train each decision
+        tree. The remainder of the data is used to fill in voting posteriors.
+
     Attributes
     ----------
     lf_ : LifelongClassificationForest
         Internal LifelongClassificationForest used to train and make
         inference.
     """
 
-    def __init__(self, n_estimators=100, finite_sample_correction=False, max_depth=30):
+    def __init__(
+        self,
+        n_estimators=100,
+        finite_sample_correction=False,
+        max_depth=30,
+        tree_construction_proportion=0.67,
+    ):
         self.n_estimators = n_estimators
         self.finite_sample_correction = finite_sample_correction
         self.max_depth = max_depth
+        self.tree_construction_proportion = tree_construction_proportion
 
     def fit(self, X, y):
         """
@@ -278,6 +289,7 @@ def fit(self, X, y):
             default_n_estimators=self.n_estimators,
             default_finite_sample_correction=self.finite_sample_correction,
             default_max_depth=self.max_depth,
+            default_tree_construction_proportion=self.tree_construction_proportion,
         )
         self.lf_.add_task(X, y, task_id=0)
         return self