added another two datasets

README.rst

@@ -114,15 +114,19 @@ Benchmarking
 Across a variety of datasets, ``XGBDistribution`` performs similarly to
 ``NGBRegressor``, but is substantially faster.
 
-+--------------+------------------------------------+-----------------------------------+------------------------+
-|              | XGBDistribution                    | NGBRegressor                      |  XGBRegressor          |
-+---------+----+-----------+-----------+------------+-----------+-----------+-----------+-----------+------------+
-| Dataset | N  | NLL       | RMSE      | Time (s)   | NLL       | RMSE      | Time (s)  | RMSE      | Time (s)   |
-+=========+====+===========+===========+============+===========+===========+===========+===========+============+
-| Boston  |506 | 2.62±0.26 | 3.41±0.69 | 0.067±0.01 | 2.55±0.24 | 3.25±0.66 | 2.68±0.45 | 3.27±0.65 | 0.035±0.01 |
-+---------+----+-----------+-----------+------------+-----------+-----------+-----------+-----------+------------+
-| Concrete|1030| 3.14±0.21 | 5.41±0.74 | 0.13±0.03  | 3.09±0.13 | 5.62±0.69 | 5.79±0.59 | 4.38±0.70 | 0.09±0.02  |
-+---------+----+-----------+-----------+------------+-----------+-----------+-----------+-----------+------------+
++---------------+------------------------------------+-----------------------------------+------------------------+
+|               | XGBDistribution                    | NGBRegressor                      |  XGBRegressor          |
++---------+-----+-----------+-----------+------------+-----------+-----------+-----------+-----------+------------+
+| Dataset | N   | NLL       | RMSE      | Time (s)   | NLL       | RMSE      | Time (s)  | RMSE      | Time (s)   |
++=========+=====+===========+===========+============+===========+===========+===========+===========+============+
+| Boston  |506  | 2.62±0.26 | 3.41±0.69 | 0.067±0.01 | 2.55±0.24 | 3.25±0.66 | 2.68±0.45 | 3.27±0.65 | 0.035±0.01 |
++---------+-----+-----------+-----------+------------+-----------+-----------+-----------+-----------+------------+
+| Concrete|1030 | 3.14±0.21 | 5.41±0.74 | 0.13±0.03  | 3.09±0.13 | 5.62±0.69 | 5.79±0.59 | 4.38±0.70 | 0.09±0.02  |
++---------+-----+-----------+-----------+------------+-----------+-----------+-----------+-----------+------------+
+| Energy  |768  | 0.58±0.41 | 0.45±0.07 | 0.15±0.03  | 0.62±0.28 | 0.49±0.07 | 5.33±0.35 | 0.40±0.06 | 0.05±0.02  |
++---------+-----+-----------+-----------+------------+-----------+-----------+-----------+-----------+------------+
+| Naval   |11934|-5.11±0.06 | 0.0014(1) | 5.8±0.85   |-3.91±0.02 | 0.0059(1) | 43.6±0.5  | 0.00123(5)| 1.93±0.07  |
++---------+-----+-----------+-----------+------------+-----------+-----------+-----------+-----------+------------+
 
 We used 10-fold cross-validation, where in each training fold 10% of the data
 were split off as a validation set, repeated over 5 random seeds. All models were

examples/benchmarking.py

@@ -226,7 +226,7 @@ def ngb_regressor(data):
 
 @evaluate
 def xgb_distribution(data):
-    xgbd = XGBDistribution(max_depth=None, natural_gradient=True, n_estimators=500)
+    xgbd = XGBDistribution(max_depth=3, natural_gradient=True, n_estimators=500)
     xgbd.fit(
         data.X_train,
         data.y_train,
@@ -239,7 +239,7 @@ def xgb_distribution(data):
 
 @evaluate
 def xgb_regressor(data):
-    xgb = XGBRegressor(max_depth=None, n_estimators=500)
+    xgb = XGBRegressor(max_depth=3, n_estimators=500)
     xgb.fit(
         data.X_train,
         data.y_train,