Unable to provide hyperparameters for Naive-Bayes and XGBoost metalearners in Stacked Ensembles

[exalate-issue-sync]

Code snippets to reproduce the bug

STEP-1
Train the base learners

{code:python}
import h2o
from h2o.estimators.random_forest import H2ORandomForestEstimator
from h2o.estimators.gbm import H2OGradientBoostingEstimator
from h2o.estimators.stackedensemble import H2OStackedEnsembleEstimator
h2o.init()

train = h2o.import_file("https://s3.amazonaws.com/erin-data/higgs/higgs_train_10k.csv")
test = h2o.import_file("https://s3.amazonaws.com/erin-data/higgs/higgs_test_5k.csv")

x = train.columns
y = "response"
x.remove(y)

train[y] = train[y].asfactor()
test[y] = test[y].asfactor()

nfolds = 5

my_gbm = H2OGradientBoostingEstimator(distribution="bernoulli",
ntrees=10,
max_depth=3,
min_rows=2,
learn_rate=0.2,
nfolds=nfolds,
fold_assignment="Modulo",
keep_cross_validation_predictions=True,
seed=1)

my_gbm.train(x=x, y=y, training_frame=train)

Train and cross-validate a RF

my_rf = H2ORandomForestEstimator(ntrees=50,
nfolds=nfolds,
fold_assignment="Modulo",
keep_cross_validation_predictions=True,
seed=1)
my_rf.train(x=x, y=y, training_frame=train)

{code}

STEP-2
Train the stacked ensemble with the naivebayes metalearner and pass params via metalearner_params.

{code:python}

ensemble = H2OStackedEnsembleEstimator(model_id="my_ensemble_binomial",
base_models=[my_gbm, my_rf],
metalearner_algorithm= 'naivebayes',
metalearner_params= {'min_prob': 0.5})

ensemble.train(x=x, y=y, training_frame=train)
{code}

The error message is as follows

{code}

H2OServerError Traceback (most recent call last)
in
4 metalearner_params= {'min_prob': 0.5})
5
----> 6 ensemble.train(x=x, y=y, training_frame=train)

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/estimators/stackedensemble.py in train(self, x, y, training_frame, blending_frame, verbose, **kwargs)
827 parms = sup._make_parms(x, y, training_frame, extend_parms_fn=extend_parms, **kwargs)
828
--> 829 sup._train(parms, verbose=verbose)

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/estimators/estimator_base.py in _train(self, parms, verbose)
192
193 rest_ver = self._get_rest_version(parms)
--> 194 model_builder_json = h2o.api("POST /%d/ModelBuilders/%s" % (rest_ver, self.algo), data=parms)
195 job = H2OJob(model_builder_json, job_type=(self.algo + " Model Build"))
196

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/h2o.py in api(endpoint, data, json, filename, save_to)
107 # type checks are performed in H2OConnection class
108 _check_connection()
--> 109 return h2oconn.request(endpoint, data=data, json=json, filename=filename, save_to=save_to)
110
111

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/backend/connection.py in request(self, endpoint, data, json, filename, save_to)
476 save_to = save_to(resp)
477 self._log_end_transaction(start_time, resp)
--> 478 return self._process_response(resp, save_to)
479
480 except (requests.exceptions.ConnectionError, requests.exceptions.HTTPError) as e:

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/backend/connection.py in _process_response(response, save_to)
827 # Note that it is possible to receive valid H2OErrorV3 object in this case, however it merely means the server
828 # did not provide the correct status code.
--> 829 raise H2OServerError("HTTP %d %s:\n%r" % (status_code, response.reason, data))
830
831

H2OServerError: HTTP 500 Server Error:
Server error java.lang.UnsupportedOperationException:
Error: Unknown meta-learner algo: naivebayes
Request: None
{code}

STEP-3
Train the stacked ensemble with the naivebayes metalearner and pass params via metalearner_params.

{code:python}

ensemble = H2OStackedEnsembleEstimator(model_id="my_ensemble_binomial",
base_models=[my_gbm, my_rf],
metalearner_algorithm= 'xgboost',
metalearner_params= {'booster': 'dart'})

ensemble.train(x=x, y=y, training_frame=train)
{code}

The corresponding error message

{code}

H2OServerError Traceback (most recent call last)
in
4 metalearner_params= {'booster': 'dart'})
5
----> 6 ensemble.train(x=x, y=y, training_frame=train)

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/estimators/stackedensemble.py in train(self, x, y, training_frame, blending_frame, verbose, **kwargs)
827 parms = sup._make_parms(x, y, training_frame, extend_parms_fn=extend_parms, **kwargs)
828
--> 829 sup._train(parms, verbose=verbose)

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/estimators/estimator_base.py in _train(self, parms, verbose)
192
193 rest_ver = self._get_rest_version(parms)
--> 194 model_builder_json = h2o.api("POST /%d/ModelBuilders/%s" % (rest_ver, self.algo), data=parms)
195 job = H2OJob(model_builder_json, job_type=(self.algo + " Model Build"))
196

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/h2o.py in api(endpoint, data, json, filename, save_to)
107 # type checks are performed in H2OConnection class
108 _check_connection()
--> 109 return h2oconn.request(endpoint, data=data, json=json, filename=filename, save_to=save_to)
110
111

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/backend/connection.py in request(self, endpoint, data, json, filename, save_to)
476 save_to = save_to(resp)
477 self._log_end_transaction(start_time, resp)
--> 478 return self._process_response(resp, save_to)
479
480 except (requests.exceptions.ConnectionError, requests.exceptions.HTTPError) as e:

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/h2o/backend/connection.py in _process_response(response, save_to)
827 # Note that it is possible to receive valid H2OErrorV3 object in this case, however it merely means the server
828 # did not provide the correct status code.
--> 829 raise H2OServerError("HTTP %d %s:\n%r" % (status_code, response.reason, data))
830
831

H2OServerError: HTTP 500 Server Error:
Server error java.lang.UnsupportedOperationException:
Error: Unknown meta-learner algo: xgboost
Request: None

{code}

[exalate-issue-sync]

Erin LeDell commented: Thanks [~accountid:557058:d55d2d14-5709-435a-be18-0436a70f4414] for the bug report. We will take a look.

[h2o-ops]

JIRA Issue Migration Info

Jira Issue: PUBDEV-7907
Assignee: Tomas Fryda
Reporter: Abhinav Sharma
State: Resolved
Fix Version: 3.34.0.4
Attachments: N/A
Development PRs: Available

Linked PRs from JIRA

#5153