update release notes, remove vi hdi report

Author: aloctavodia

RELEASE-NOTES.md

@@ -14,8 +14,9 @@
 ### New features
 - `sample_posterior_predictive_w` can now feed on `xarray.Dataset` - e.g. from `InferenceData.posterior`. (see [#4042](https://github.com/pymc-devs/pymc3/pull/4042))
 - Add MLDA, a new stepper for multilevel sampling. MLDA can be used when a hierarchy of approximate posteriors of varying accuracy is available, offering improved sampling efficiency especially in high-dimensional problems and/or where gradients are not available (see [#3926](https://github.com/pymc-devs/pymc3/pull/3926))
-- Change SMC metropolis kernel to independent metropolis kernel [#4115](https://github.com/pymc-devs/pymc3/pull/3926))
+- Change SMC metropolis kernel to independent metropolis kernel [#4115](https://github.com/pymc-devs/pymc3/pull/4115))
 - Add alternative parametrization to NegativeBinomial distribution in terms of n and p (see [#4126](https://github.com/pymc-devs/pymc3/issues/4126))
+- Add Bayesian Additive Regression Trees (BARTs) [#4183](https://github.com/pymc-devs/pymc3/pull/4183))
 
 
 ## PyMC3 3.9.3 (11 August 2020)

pymc3/sampling.py

@@ -607,13 +607,8 @@ def sample(
     trace.report._t_sampling = t_sampling
 
     if "variable_inclusion" in trace.stat_names:
-        variable_inclusion = np.vstack(trace.get_sampler_stats("variable_inclusion"))
-        variable_inclusion = np.vstack(
-            [v.sum(0) / v.sum() for v in np.array_split(variable_inclusion, 50)]
-        )
-        trace.report.variable_importance = np.empty((variable_inclusion.shape[1], 3))
-        trace.report.variable_importance[:, 0] = variable_inclusion.mean(0)
-        trace.report.variable_importance[:, 1:3] = arviz.hdi(variable_inclusion, hdi_prob=0.68)
+        variable_inclusion = np.stack(trace.get_sampler_stats("variable_inclusion")).mean(0)
+        trace.report.variable_importance = variable_inclusion / variable_inclusion.sum()
 
     n_chains = len(trace.chains)
     _log.info(

pymc3/step_methods/pgbart.py

@@ -12,6 +12,8 @@
 #   See the License for the specific language governing permissions and
 #   limitations under the License.
 
+import logging
+
 import numpy as np
 from theano import function as theano_function
 
@@ -21,6 +23,8 @@
 from ..model import modelcontext
 from ..theanof import inputvars, make_shared_replacements, join_nonshared_inputs
 
+_log = logging.getLogger("pymc3")
+
 
 class PGBART(ArrayStepShared):
     """
@@ -52,6 +56,7 @@ class PGBART(ArrayStepShared):
     stats_dtypes = [{"variable_inclusion": np.ndarray}]
 
     def __init__(self, vars=None, num_particles=10, max_stages=5000, chunk="auto", model=None):
+        _log.warning("The BART model is experimental. Use with caution.")
         model = modelcontext(model)
         vars = inputvars(vars)
         self.bart = vars[0].distribution