Merge pull request #22 from lacava/issues_fix

changes for parents pressure for selection

few/few.py

@@ -59,7 +59,7 @@ def __init__(self, population_size=50, generations=100,
                  random_state=None, verbosity=0,
                  scoring_function=None, disable_update_check=False,
                  elitism=True, boolean = False,classification=False,clean=False,
-                 track_diversity=False,mdr=False,otype='f',c=True):
+                 track_diversity=False,mdr=False,otype='f',c=True, weight_parents=False):
                 # sets up GP.
 
         # Save params to be recalled later by get_params()
@@ -93,6 +93,7 @@ def __init__(self, population_size=50, generations=100,
         self.fit_choice = fit_choice
         self.op_weight = op_weight
         self.max_stall = max_stall
+        self.weight_parents = weight_parents
         self.seed_with_ml = seed_with_ml
         self.erc = erc
         self.random_state = check_random_state(random_state)
@@ -754,6 +755,9 @@ def main():
                         type=positive_integer, help='If model CV does not '
                         'improve for this many generations, end optimization.')
 
+    parser.add_argument('--weight_parents', action='store_true',dest='WEIGHT_PARENTS',default=False,
+                        help='Feature importance determines parent pressure for selection.')
+
     parser.add_argument('-sel', action='store', dest='SEL',
                         default='epsilon_lexicase',
                         choices = ['tournament','lexicase','epsilon_lexicase',
@@ -881,7 +885,7 @@ def main():
                   fit_choice = args.FIT_CHOICE,boolean=args.BOOLEAN,
                   classification=args.CLASSIFICATION,clean = args.CLEAN,
                   track_diversity=args.TRACK_DIVERSITY,mdr=args.MDR,
-                  otype=args.OTYPE,c=args.c)
+                  otype=args.OTYPE,c=args.c, weight_parents = args.WEIGHT_PARENTS)
 
     learner.fit(training_features, training_labels)
     # pdb.set_trace()

few/tests/test_few.py

@@ -47,6 +47,30 @@ def test_few_fit_shapes():
     "test r2:",r2_score(boston.target[300:],yhat_test))
     assert yhat_test.shape == boston.target[300:].shape
 
+def test_few_with_parents_weight():
+    """test_few.py: few performs without error with parent pressure for selection"""
+    np.random.seed(1006987)
+    boston = load_boston()
+    d = np.column_stack((boston.data,boston.target))
+    np.random.shuffle(d)
+    features = d[:,0:-1]
+    target = d[:,-1]
+
+    print("feature shape:",boston.data.shape)
+
+    learner = FEW(generations=1, population_size=5,
+                mutation_rate=1, crossover_rate=1,
+                ml = LassoLarsCV(), min_depth = 1, max_depth = 3,
+                sel = 'tournament', fit_choice = 'r2',tourn_size = 2, random_state=0, verbosity=0,
+                disable_update_check=False, weight_parents=True)
+
+    learner.fit(features[:300], target[:300])
+    few_score = learner.score(features[:300], target[:300])
+    test_score = learner.score(features[300:],target[300:])
+
+    print("few score:",few_score)
+    print("few test score:",test_score)
+
 
 def test_few_at_least_as_good_as_default():
     """test_few.py: few performs at least as well as the default ML """

few/variation.py

@@ -22,10 +22,20 @@ def variation(self,parents):
         if type(self.ml).__name__ != 'SVC' and type(self.ml).__name__ != 'SVR': # this is needed because svm has a bug that throws valueerror on attribute check
             if hasattr(self.ml,'coef_'):
                 # for l1 regularization, filter individuals with 0 coefficients
-                offspring = copy.deepcopy(list(x for i,x in zip(self.ml.coef_, self.valid(parents)) if  (i != 0).any()))
+                if self.weight_parents:
+                    weights = abs(self.ml.coef_)
+                    weights = weights/sum(weights)
+                    offspring = copy.deepcopy(list(np.random.choice(self.valid(parents), self.population_size, p=weights)))
+                else:
+                    offspring = copy.deepcopy(list(x for i,x in zip(self.ml.coef_, self.valid(parents)) if  (i != 0).any()))
             elif hasattr(self.ml,'feature_importances_'):
                 # for tree methods, filter our individuals with 0 feature importance
-                offspring = copy.deepcopy(list(x for i,x in zip(self.ml.feature_importances_, self.valid(parents)) if  i != 0))
+                if self.weight_parents:
+                    weights = self.ml.feature_importances_
+                    weights = weights/sum(weights)
+                    offspring = copy.deepcopy(list(np.random.choice(self.valid(parents), self.population_size, p=weights)))
+                else:
+                    offspring = copy.deepcopy(list(x for i,x in zip(self.ml.feature_importances_, self.valid(parents)) if  i != 0))
             else:
                 offspring = copy.deepcopy(self.valid(parents))
         else: