move feature usage to hypergbm

DataCanvasIO · May 5, 2023 · ededd1c · ededd1c
1 parent f73040d
commit ededd1c
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 17 deletions.
diff --git a/hypernets/model/objectives.py b/hypernets/model/objectives.py
@@ -13,6 +13,7 @@
 
 random_state = get_random_state()
 
+
 def calc_psi(x_array, y_array, n_bins=10, eps=1e-6):
  def calc_ratio(y_proba):
  y_proba_1d = y_proba.reshape(1, -1)
@@ -131,18 +132,6 @@ def _evaluate_cv(self, trial, estimator, X_trains, y_trains, X_vals, y_vals, X_t
  return self._get_psi_score(estimator, X_train=X_train, X_test=X_test)
 
 
-class FeatureUsageObjective(Objective):
- def __init__(self):
- super(FeatureUsageObjective, self).__init__('feature_usage', 'min', need_train_data=False,
- need_val_data=True, need_test_data=False)
-
- def _evaluate(self, trial, estimator, X_train, y_train, X_val, y_val, X_test=None, **kwargs) -> float:
- return estimator.data_pipeline[0].features[0][1].steps[0][1].feature_usage()
-
- def _evaluate_cv(self, trial, estimator, X_trains, y_trains, X_vals, y_vals, X_test=None, **kwargs) -> float:
- return estimator.cv_models_[0].data_pipeline[0].features[0][1].steps[0][1].feature_usage()
-
-
 class ElapsedObjective(Objective):
 
  def __init__(self):
@@ -368,8 +357,6 @@ def copy_opt(opt_names):
  elif name == 'psi':
  copy_opt(['n_bins', 'task', 'average', 'eps'])
  return PSIObjective(**opts)
- elif name == 'feature_usage':
- return FeatureUsageObjective()
  elif name == 'pred_perf':
  return PredictionPerformanceObjective()
  else:

diff --git a/hypernets/tests/searchers/test_nsga2_searcher.py b/hypernets/tests/searchers/test_nsga2_searcher.py
@@ -137,10 +137,9 @@ class TestRNSGA2:
 
  @pytest.mark.parametrize('recombination', ["shuffle", "uniform", "single_point"])
  @pytest.mark.parametrize('cv', [True, False])
- @pytest.mark.parametrize('objective', ['feature_usage', 'nf'])
- def test_nsga2_training(self, recombination: str, cv: bool, objective: str):
+ def test_nsga2_training(self, recombination: str, cv: bool):
  set_random_state(1234)
- hk1 = self.run_nsga2_training(recombination=const.COMBINATION_SHUFFLE, cv=cv, objective=objective)
+ hk1 = self.run_nsga2_training(recombination=const.COMBINATION_SHUFFLE, cv=cv, objective='nf')
  pop1 = hk1.searcher.get_historical_population()
  scores1 = np.asarray([indi.scores for indi in pop1])
  assert scores1.ndim == 2