Merge pull request #42 from GauravPandeyLab/minor-fixes

Minor fixes

eipy/ei.py

@@ -31,6 +31,8 @@
 from eipy.metrics import (
  base_summary,
  ensemble_summary,
+ fmax_score,
+ roc_auc_score
 )
 
 warnings.filterwarnings("ignore", category=DeprecationWarning)
@@ -70,8 +72,7 @@ class EnsembleIntegration:
  n_jobs : int, default=1
  Number of workers for parallelization in joblib.
  metrics : dict, default=None
- If None, the maximized F1-score and AUC scores are calculated. If specified, keys ending
- in 'max' or 'min' will maximize/minimize the given metric by calculating a threshold.
+ If None, the maximized F1-score and AUC scores are calculated.
  random_state : int, default=None
  Random state for cross-validation and use in some models.
  parallel_backend : str, default='loky'

eipy/interpretation.py

@@ -10,6 +10,7 @@
 from operator import itemgetter
 from sklearn.ensemble import VotingClassifier
 from sklearn.preprocessing import LabelEncoder
+from eipy.metrics import fmax_score
 
 import warnings
 
@@ -25,7 +26,7 @@ class PermutationInterpreter:
  EI : EnsembleIntegration class object
  Fitted EnsembleIntegration model, i.e. with model_building=True.
  metric : function
- sklearn-like metric function.
+ sklearn-like metric function. If None, the fmax score is used.
  n_repeats : int, default=10
  Number of repeats in PermutationImportance.
  ensemble_predictor_keys: default='all'
@@ -54,14 +55,19 @@ class PermutationInterpreter:
  def __init__(
  self,
  EI,
- metric,
+ metric=None,
  ensemble_predictor_keys="all", # can be "all" or a list of keys for ensemble methods
  n_repeats=10,
  n_jobs=1,
  metric_greater_is_better=True,
  ):
  self.EI = EI
- self.metric = metric
+
+ if metric is None: # use fmax score if metric not specified
+ self.metric = lambda y_test, y_pred: fmax_score(y_test, y_pred)[0]
+ else:
+ self.metric = metric
+
  self.n_repeats = n_repeats
  self.n_jobs = n_jobs
  self.ensemble_predictor_keys = ensemble_predictor_keys