From 2b978a8067b850ed6556bdc4fffd4732b8c263b0 Mon Sep 17 00:00:00 2001
From: Baptiste Bauvin <baptiste.bauvin@lis-lab.fr>
Date: Thu, 13 Dec 2018 13:59:45 -0500
Subject: [PATCH] Added data extraction
---
.../Monoview/Additions/BoostUtils.py | 14 ++-
.../Monoview/Additions/QarBoostUtils.py | 91 ++++++++++++++-----
.../MonoviewClassifiers/QarBoostNC.py | 5 +-
.../MonoviewClassifiers/QarBoostNC2.py | 15 ++-
.../MonoviewClassifiers/QarBoostNC3.py | 14 +--
.../MonoviewClassifiers/QarBoostv2.py | 13 +--
6 files changed, 95 insertions(+), 57 deletions(-)
diff --git a/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/BoostUtils.py b/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/BoostUtils.py
index f5191941..b1eb2540 100644
--- a/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/BoostUtils.py
+++ b/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/BoostUtils.py
@@ -669,7 +669,7 @@ class ConvexProgram(object):
return signs
-def get_accuracy_graph(train_accuracies, classifier_name, file_name, name="Accuracies", bounds=None):
+def get_accuracy_graph(train_accuracies, classifier_name, file_name, name="Accuracies", bounds=None, bound_name=None, boosting_bound=None):
if type(name) is not str:
name = " ".join(name.getConfig().strip().split(" ")[:2])
if bounds:
@@ -677,8 +677,14 @@ def get_accuracy_graph(train_accuracies, classifier_name, file_name, name="Accur
ax.set_title(name+" during train for "+classifier_name)
x = np.arange(len(train_accuracies))
scat = ax.scatter(x, np.array(train_accuracies), )
- scat2 = ax.scatter(x, np.array(bounds), )
- ax.legend((scat,scat2), (name,"Bounds"))
+ if boosting_bound:
+ scat2 = ax.scatter(x, boosting_bound)
+ scat3 = ax.scatter(x, np.array(bounds), )
+ ax.legend((scat, scat2, scat3), (name,"Boosting bound", bound_name))
+ else:
+ scat2 = ax.scatter(x, np.array(bounds), )
+ ax.legend((scat, scat2),
+ (name, bound_name))
plt.tight_layout()
f.savefig(file_name)
plt.close()
@@ -751,5 +757,5 @@ def getInterpretBase(classifier, directory, classifier_name, weights,
separator=',', suppress_small=True)
np.savetxt(directory + "voters.csv", classifier.classification_matrix[:, classifier.chosen_columns_], delimiter=',')
np.savetxt(directory + "weights.csv", classifier.weights_, delimiter=',')
- get_accuracy_graph(classifier.train_metrics, classifier_name, directory + 'metrics.png', classifier.plotted_metric, classifier.bounds)
+ get_accuracy_graph(classifier.train_metrics, classifier_name, directory + 'metrics.png', classifier.plotted_metric, classifier.bounds, "Boosting bound")
return interpretString
diff --git a/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/QarBoostUtils.py b/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/QarBoostUtils.py
index 65c4abb4..ee475b02 100644
--- a/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/QarBoostUtils.py
+++ b/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/QarBoostUtils.py
@@ -18,8 +18,8 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
def __init__(self, n_max_iterations=None, estimators_generator=None,
random_state=42, self_complemented=True, twice_the_same=False,
c_bound_choice=True, random_start=True,
- n_stumps_per_attribute=None, use_r=True,
- plotted_metric=Metrics.zero_one_loss):
+ n_stumps_per_attribute=None, use_r=True, c_bound_sol=True,
+ plotted_metric=Metrics.zero_one_loss, save_train_data=True):
super(ColumnGenerationClassifierQar, self).__init__()
r"""
@@ -60,10 +60,12 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
if n_stumps_per_attribute:
self.n_stumps = n_stumps_per_attribute
self.use_r = use_r
+ self.c_bound_sol = c_bound_sol
+ self.save_train_data = save_train_data
self.printed_args_name_list = ["n_max_iterations", "self_complemented",
"twice_the_same",
"c_bound_choice", "random_start",
- "n_stumps", "use_r"]
+ "n_stumps", "use_r", "c_bound_sol"]
def set_params(self, **params):
self.self_complemented = params["self_complemented"]
@@ -94,7 +96,8 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
# Print dynamically the step and the error of the current classifier
print(
- "Resp. bound : {}, {}/{}, eps :{}".format(self.respected_bound,
+ "Resp. bound : {}, {}; {}/{}, eps :{}".format(self.respected_bound,
+ self.bounds[-1] > self.train_metrics[-1],
k + 2,
self.n_max_iterations,
self.voter_perfs[-1]),
@@ -111,7 +114,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
voter_perf = self.compute_voter_perf(formatted_y)
- self.compute_voter_weight(voter_perf)
+ self.compute_voter_weight(voter_perf, sol)
self.update_example_weights(formatted_y)
@@ -121,6 +124,10 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
self.estimators_generator.estimators_ = \
self.estimators_generator.estimators_[self.chosen_columns_]
+ if self.save_train_data:
+ self.X_train = self.classification_matrix[:, self.chosen_columns_]
+ self.y_train = formatted_y
+
self.weights_ = np.array(self.weights_)
self.weights_ /= np.sum(self.weights_)
@@ -170,12 +177,15 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
self.train_metrics.append(train_metric)
self.bounds.append(bound)
- def compute_voter_weight(self, voter_perf):
+ def compute_voter_weight(self, voter_perf, sol):
"""used to compute the voter's weight according to the specified method (edge or error) """
- if self.use_r:
- self.q = 0.5 * math.log((1 + voter_perf) / (1 - voter_perf))
+ if self.c_bound_sol:
+ self.q = sol
else:
- self.q = math.log((1 - voter_perf) / voter_perf)
+ if self.use_r:
+ self.q = 0.5 * math.log((1 + voter_perf) / (1 - voter_perf))
+ else:
+ self.q = math.log((1 - voter_perf) / voter_perf)
self.weights_.append(self.q)
def compute_voter_perf(self, formatted_y):
@@ -230,11 +240,13 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
else:
epsilon = self._compute_epsilon(y)
self.voter_perfs.append(epsilon)
-
- if self.use_r:
- self.q = 0.5 * math.log((1 + r) / (1 - r))
+ if self.c_bound_sol:
+ self.q = 1
else:
- self.q = math.log((1 - epsilon) / epsilon)
+ if self.use_r:
+ self.q = 0.5 * math.log((1 + r) / (1 - r))
+ else:
+ self.q = math.log((1 - epsilon) / epsilon)
self.weights_.append(self.q)
# Update the distribution on the examples.
@@ -290,6 +302,8 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
self.example_weights_ = []
self.train_metrics = []
self.bounds = []
+ self.disagreements = []
+ self.margins = []
self.previous_votes = []
self.previous_margins = []
self.respected_bound = True
@@ -373,11 +387,13 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
if not causes:
causes = ["no feature was better than random and acceptable"]
if c_borns:
- min_c_born_index = ma.argmin(c_borns)
- self.c_bounds.append(c_borns[min_c_born_index])
- selected_sol = possible_sols[min_c_born_index]
- selected_voter_index = indices[min_c_born_index]
- return selected_sol, selected_voter_index
+ min_c_bound_index = ma.argmin(c_borns)
+ self.c_bounds.append(c_borns[min_c_bound_index])
+ selected_sol = possible_sols[min_c_bound_index]
+ self.margins.append(self.margin(selected_sol))
+ self.disagreements.append(self.disagreement(selected_sol))
+ selected_voter_index = indices[min_c_bound_index]
+ return selected_sol/(1+selected_sol), selected_voter_index
else:
return "break", " and ".join(set(causes))
@@ -388,7 +404,10 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
zero_diag = np.ones((m, m)) - np.identity(m)
weighted_previous_sum = np.multiply(y,
self.previous_vote.reshape((m, 1)))
- weighted_next_column = np.multiply(next_column.reshape((m, 1)),
+ if self.c_bound_sol:
+ weighted_next_column = next_column.reshape((m, 1))
+ else:
+ weighted_next_column = np.multiply(next_column.reshape((m, 1)),
self.example_weights.reshape((m, 1)))
self.B2 = np.sum(weighted_next_column ** 2)
@@ -437,12 +456,17 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
""""We just check that the solution found by np.roots is acceptable under our constraints
(real, a minimum and over 0)"""
if sols.shape[0] == 1:
- if self._cbound(sols[0]) < self._cbound(sols[0] + 1):
+ if self._cbound(sols[0]) < self._cbound(sols[0] + 1) and sols[0] > 0:
best_sol = sols[0]
else:
- return False, "the only solution was a maximum."
- elif sols.shape[0] == 2:
+ if sols[0] > 0:
+ return False, "the only solution was a maximum."
+ else:
+ return False, "the only solution was negative"
+ elif sols.shape[0] == 2 and sols[0] > 0 and sols[1] > 1:
best_sol = self._best_sol(sols)
+ elif np.greater(sols, np.zeros(2)).any():
+ return self._analyze_solutions_one_weight(np.array([np.max(sols)]))
else:
return False, "no solution were found"
@@ -453,8 +477,14 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
def _cbound(self, sol):
"""Computing the objective function"""
- return 1 - (self.A2 * sol ** 2 + self.A1 * sol + self.A0) / (
- self.B2 * sol ** 2 + self.B1 * sol + self.B0) / self.n_total_examples
+ return 1 - (self.A2 * sol ** 2 + self.A1 * sol + self.A0) / ((
+ self.B2 * sol ** 2 + self.B1 * sol + self.B0) * self.n_total_examples)
+
+ def disagreement(self, sol):
+ return self.B2 * sol ** 2 + self.B1 * sol + self.B0
+
+ def margin(self, sol):
+ return (self.A2 * sol ** 2 + self.A1 * sol + self.A0)/self.n_total_examples
def _best_sol(self, sols):
"""Return the best min in the two possible sols"""
@@ -501,9 +531,20 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
shutil.rmtree(path + "/gif_images")
get_accuracy_graph(self.voter_perfs, self.__class__.__name__,
directory + 'voter_perfs.png', "Errors")
+ get_accuracy_graph(self.c_bounds, self.__class__.__name__,
+ directory + 'c_bounds.png', "C-Bounds")
+ get_accuracy_graph(self.margins, self.__class__.__name__,
+ directory + 'margins.png', "Margins")
+ self.disagreements[0] = 0
+ get_accuracy_graph(self.disagreements, self.__class__.__name__,
+ directory + 'disagreements.png', "disagreements")
+ get_accuracy_graph(self.train_metrics[1:], self.__class__.__name__,
+ directory + 'c_bounds_train_metrics.png', self.plotted_metric, self.c_bounds, "C-Bound", self.bounds[1:])
interpretString = getInterpretBase(self, directory, "QarBoost",
self.weights_, self.break_cause)
-
+ if self.save_train_data:
+ np.savetxt(directory+"x_train.csv", self.X_train, delimiter=',')
+ np.savetxt(directory+"y_train.csv", self.y_train, delimiter=',')
args_dict = dict(
(arg_name, str(self.__dict__[arg_name])) for arg_name in
self.printed_args_name_list)
diff --git a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC.py b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC.py
index 97197e80..6bf35c91 100644
--- a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC.py
+++ b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC.py
@@ -6,14 +6,15 @@ from ..Monoview.Additions.QarBoostUtils import ColumnGenerationClassifierQar
class QarBoostNC(ColumnGenerationClassifierQar, BaseMonoviewClassifier):
def __init__(self, random_state=None, **kwargs):
- super(QarBoostNC, self).__init__(n_max_iterations=50,
+ super(QarBoostNC, self).__init__(n_max_iterations=500,
random_state=random_state,
self_complemented=True,
twice_the_same=False,
c_bound_choice=True,
random_start=False,
n_stumps_per_attribute=1,
- use_r=True
+ use_r=True,
+ c_bound_sol=False
)
self.param_names = []
self.distribs = []
diff --git a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC2.py b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC2.py
index 2739affe..e27a8a4d 100644
--- a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC2.py
+++ b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC2.py
@@ -6,18 +6,15 @@ from ..Monoview.Additions.QarBoostUtils import ColumnGenerationClassifierQar
class QarBoostNC2(ColumnGenerationClassifierQar, BaseMonoviewClassifier):
def __init__(self, random_state=None, **kwargs):
- super(QarBoostNC2, self).__init__(
+ super(QarBoostNC2, self).__init__(n_max_iterations=500,
random_state=random_state,
self_complemented=True,
- twice_the_same=False,
- old_fashioned=False,
- previous_vote_weighted=False,
+ twice_the_same=True,
c_bound_choice=True,
- random_start=True,
- two_wieghts_problem=False,
- divided_ponderation=False,
- n_stumps_per_attribute=10,
- use_r=True
+ random_start=False,
+ n_stumps_per_attribute=1,
+ use_r=True,
+ c_bound_sol=False
)
self.param_names = []
self.distribs = []
diff --git a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC3.py b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC3.py
index 7cb49df5..1c44cc3b 100644
--- a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC3.py
+++ b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC3.py
@@ -6,19 +6,15 @@ from ..Monoview.Additions.QarBoostUtils import ColumnGenerationClassifierQar
class QarBoostNC3(ColumnGenerationClassifierQar, BaseMonoviewClassifier):
def __init__(self, random_state=None, **kwargs):
- super(QarBoostNC3, self).__init__(
+ super(QarBoostNC3, self).__init__(n_max_iterations=500,
random_state=random_state,
- self_complemented=False,
+ self_complemented=True,
twice_the_same=False,
- old_fashioned=False,
- previous_vote_weighted=False,
c_bound_choice=True,
- random_start=True,
- two_wieghts_problem=False,
- divided_ponderation=True,
+ random_start=False,
n_stumps_per_attribute=1,
- use_r=True
- )
+ use_r=True,
+ c_bound_sol=True)
self.param_names = []
self.distribs = []
self.classed_params = []
diff --git a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostv2.py b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostv2.py
index 7fb2239a..e06e0698 100644
--- a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostv2.py
+++ b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostv2.py
@@ -6,18 +6,15 @@ from ..Monoview.Additions.QarBoostUtils import ColumnGenerationClassifierQar
class QarBoostv2(ColumnGenerationClassifierQar, BaseMonoviewClassifier):
def __init__(self, random_state=None, **kwargs):
- super(QarBoostv2, self).__init__(
+ super(QarBoostv2, self).__init__(n_max_iterations=500,
random_state=random_state,
- self_complemented=False,
- twice_the_same=False,
- old_fashioned=False,
- previous_vote_weighted=False,
+ self_complemented=True,
+ twice_the_same=True,
c_bound_choice=True,
random_start=False,
- two_wieghts_problem=False,
- divided_ponderation=False,
n_stumps_per_attribute=1,
- use_r=True
+ use_r=True,
+ c_bound_sol=True
)
self.param_names = []
self.distribs = []
--
GitLab