diff --git a/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/BoostUtils.py b/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/BoostUtils.py
index b1eb25405df720c972276146ebdcb4e4120a6f28..503bae21b04262b51e873e6d3a09b0d3029c9dcb 100644
--- a/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/BoostUtils.py
+++ b/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/BoostUtils.py
@@ -676,23 +676,23 @@ def get_accuracy_graph(train_accuracies, classifier_name, file_name, name="Accur
f, ax = plt.subplots(nrows=1, ncols=1)
ax.set_title(name+" during train for "+classifier_name)
x = np.arange(len(train_accuracies))
- scat = ax.scatter(x, np.array(train_accuracies), )
+ scat = ax.scatter(x, np.array(train_accuracies), marker=".")
if boosting_bound:
- scat2 = ax.scatter(x, boosting_bound)
- scat3 = ax.scatter(x, np.array(bounds), )
+ scat2 = ax.scatter(x, boosting_bound, marker=".")
+ scat3 = ax.scatter(x, np.array(bounds), marker=".", )
ax.legend((scat, scat2, scat3), (name,"Boosting bound", bound_name))
else:
- scat2 = ax.scatter(x, np.array(bounds), )
+ scat2 = ax.scatter(x, np.array(bounds), marker=".", )
ax.legend((scat, scat2),
(name, bound_name))
- plt.tight_layout()
+ # plt.tight_layout()
f.savefig(file_name)
plt.close()
else:
f, ax = plt.subplots(nrows=1, ncols=1)
ax.set_title(name+" during train for "+classifier_name)
x = np.arange(len(train_accuracies))
- scat = ax.scatter(x, np.array(train_accuracies), )
+ scat = ax.scatter(x, np.array(train_accuracies), marker=".", )
ax.legend((scat,), (name,))
plt.tight_layout()
f.savefig(file_name)
diff --git a/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/QarBoostUtils.py b/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/QarBoostUtils.py
index ee475b02b7fa344638912235a2a82793b0bb8179..8566f3e3ef737e4b079b3b33414723c211517acb 100644
--- a/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/QarBoostUtils.py
+++ b/multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/QarBoostUtils.py
@@ -19,7 +19,8 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
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, c_bound_sol=True,
- plotted_metric=Metrics.zero_one_loss, save_train_data=True):
+ plotted_metric=Metrics.zero_one_loss, save_train_data=True,
+ test_graph=True):
super(ColumnGenerationClassifierQar, self).__init__()
r"""
@@ -62,6 +63,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
self.use_r = use_r
self.c_bound_sol = c_bound_sol
self.save_train_data = save_train_data
+ self.test_graph = test_graph
self.printed_args_name_list = ["n_max_iterations", "self_complemented",
"twice_the_same",
"c_bound_choice", "random_start",
@@ -102,7 +104,6 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
self.n_max_iterations,
self.voter_perfs[-1]),
end="\r")
-
sol, new_voter_index = self.choose_new_voter(y_kernel_matrix,
formatted_y)
@@ -145,6 +146,8 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
logging.warning('Converting sparse matrix to dense matrix.')
X = np.array(X.todense())
classification_matrix = self._binary_classification_matrix(X)
+ # if self.test_graph:
+ # self.make_test_graph(classification_matrix)
margins = np.squeeze(
np.asarray(np.matmul(classification_matrix, self.weights_)))
signs_array = np.array([int(x) for x in sign(margins)])
@@ -164,6 +167,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
self.previous_margins.append(
np.multiply(y, self.previous_vote))
+ self.selected_margins.append(np.sum(np.multiply(y, self.new_voter)))
train_metric = self.plotted_metric.score(y, np.sign(self.previous_vote))
if self.use_r:
bound = self.bounds[-1] * math.sqrt(1 - voter_perf ** 2)
@@ -187,6 +191,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
else:
self.q = math.log((1 - voter_perf) / voter_perf)
self.weights_.append(self.q)
+ # self.weights_ = [weight/(1.0*sum(self.weights_)) for weight in self.weights_]
def compute_voter_perf(self, formatted_y):
"""Used to computer the performance (error or edge) of the selected voter"""
@@ -219,7 +224,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
"""THis initialization corressponds to the first round of boosting with equal weights for each examples and the voter chosen by it's margin."""
self.example_weights = self._initialize_alphas(m).reshape((m, 1))
- self.previous_margins.append(np.multiply(y, y))
+ # self.previous_margins.append(np.multiply(y, y))
self.example_weights_.append(self.example_weights)
if self.random_start:
first_voter_index = self.random_state.choice(
@@ -255,6 +260,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
self.previous_margins.append(
np.multiply(y, self.previous_vote))
+ self.selected_margins.append(np.sum(np.multiply(y, self.previous_vote)))
train_metric = self.plotted_metric.score(y, np.sign(self.previous_vote))
if self.use_r:
@@ -307,6 +313,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
self.previous_votes = []
self.previous_margins = []
self.respected_bound = True
+ self.selected_margins = []
def _compute_epsilon(self, y):
"""Updating the error variable, the old fashioned way uses the whole majority vote to update the error"""
@@ -318,6 +325,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
def _compute_r(self, y):
ones_matrix = np.ones(y.shape)
+
ones_matrix[np.multiply(y, self.new_voter.reshape(
y.shape)) < 0] = -1 # can np.divide if needed
r = np.average(ones_matrix, weights=self.example_weights, axis=0)
@@ -340,14 +348,15 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
pseudo_h_values = ma.array(np.sum(weighted_kernel_matrix, axis=0),
fill_value=-np.inf)
pseudo_h_values[self.chosen_columns_] = ma.masked
- acceptable_indices = np.where(
- np.logical_and(np.greater(upper_bound, pseudo_h_values),
- np.greater(pseudo_h_values, lower_bound)))[0]
- if acceptable_indices.size > 0:
- worst_h_index = self.random_state.choice(acceptable_indices)
- return worst_h_index, [0]
- else:
- return " no margin over random and acceptable", ""
+ return np.argmax(pseudo_h_values), [0]
+ # acceptable_indices = np.where(
+ # np.logical_and(np.greater(upper_bound, pseudo_h_values),
+ # np.greater(pseudo_h_values, lower_bound)))[0]
+ # if acceptable_indices.size > 0:
+ # worst_h_index = self.random_state.choice(acceptable_indices)
+ # return worst_h_index, [0]
+ # else:
+ # return " no margin over random and acceptable", ""
def _is_not_too_wrong(self, hypothese, y):
"""Check if the weighted margin is better than random"""
@@ -367,7 +376,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
def _find_new_voter(self, y_kernel_matrix, y):
"""Here, we solve the two_voters_mincq_problem for each potential new voter,
and select the one that has the smallest minimum"""
- c_borns = []
+ c_bounds = []
possible_sols = []
indices = []
causes = []
@@ -379,21 +388,21 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
and self._is_not_too_wrong(hypothese, y):
w = self._solve_one_weight_min_c(hypothese, y)
if w[0] != "break":
- c_borns.append(self._cbound(w[0]))
+ c_bounds.append(self._cbound(w[0]))
possible_sols.append(w)
indices.append(hypothese_index)
else:
causes.append(w[1])
if not causes:
causes = ["no feature was better than random and acceptable"]
- if c_borns:
- min_c_bound_index = ma.argmin(c_borns)
- self.c_bounds.append(c_borns[min_c_bound_index])
+ if c_bounds:
+ min_c_bound_index = ma.argmin(c_bounds)
+ self.c_bounds.append(c_bounds[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
+ return selected_sol, selected_voter_index #selected_sol/(1+selected_sol)
else:
return "break", " and ".join(set(causes))
@@ -441,10 +450,18 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
np.array(float(C0) / C1).reshape((1, 1)))
if is_acceptable:
return np.array([sol])
- try:
- sols = np.roots(np.array([C2, C1, C0]))
- except:
- return ["break", "nan"]
+ if C1*C1 < 4*C2*C0:
+ return ['break', "no roots"]
+ if C2 > 0 and C1*C1 == 4*C2*C0:
+ return ['break', "maximum"]
+
+ if C1*C1-4*C2*C0<=0:
+ try:
+ sols = np.roots(np.array([C2, C1, C0]))
+ except:
+ return ["break", "nan"]
+ else:
+ sols = np.array([(-C1-math.sqrt(C1*C1-4*C2*C0))/(2*C2), (-C1+math.sqrt(C1*C1-4*C2*C0))/(2*C2)])
is_acceptable, sol = self._analyze_solutions_one_weight(sols)
if is_acceptable:
@@ -481,7 +498,7 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
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
+ return (self.B2 * sol ** 2 + self.B1 * sol + self.B0)/self.n_total_examples
def margin(self, sol):
return (self.A2 * sol ** 2 + self.A1 * sol + self.A0)/self.n_total_examples
@@ -495,51 +512,70 @@ class ColumnGenerationClassifierQar(BaseEstimator, ClassifierMixin, BaseBoost):
"""Initialize the examples wieghts"""
return 1.0 / n_examples * np.ones((n_examples,))
+ # def make_test_graph(self, classification_matrix):
+ # signs_arrays = np.array([self.predict_classification_matrix(classification_matrix, index+1) for index in range(self.weights_.shape[0])]).reshape((classification_matrix.shape))
+ # np.savetxt("/home/baptiste/signs_arrays.csv",
+ # signs_arrays.astype(np.int16), delimiter=",")
+ #
+ # def predict_classification_matrix(self, classification_matrix, index):
+ # margins = np.squeeze(
+ # np.asarray(np.matmul(classification_matrix[:, :index], self.weights_[:index])))
+ # signs_array = np.array([int(x) for x in sign(margins)])
+ # signs_array[signs_array == -1] = 0
+ # return signs_array
+
def getInterpretQar(self, directory):
+ self.directory = directory
"""Used to interpret the functionning of the algorithm"""
path = "/".join(directory.split("/")[:-1])
- try:
- import os
- os.makedirs(path + "/gif_images")
- except:
- raise
- filenames = []
- max_weight = max([np.max(examples_weights) for examples_weights in
- self.example_weights_])
- min_weight = min([np.max(examples_weights) for examples_weights in
- self.example_weights_])
- for iterIndex, examples_weights in enumerate(self.example_weights_):
- r = np.array(examples_weights)
- theta = np.arange(self.n_total_examples)
- colors = np.sign(self.previous_margins[iterIndex])
- fig = plt.figure(figsize=(5, 5), dpi=80)
- ax = fig.add_subplot(111)
- c = ax.scatter(theta, r, c=colors, cmap='RdYlGn', alpha=0.75)
- ax.set_ylim(min_weight, max_weight)
- filename = path + "/gif_images/" + str(iterIndex) + ".png"
- filenames.append(filename)
- plt.savefig(filename)
- plt.close()
-
- import imageio
- images = []
- logging.getLogger("PIL").setLevel(logging.WARNING)
- for filename in filenames:
- images.append(imageio.imread(filename))
- imageio.mimsave(path + '/weights.gif', images, duration=1. / 2)
- import shutil
- shutil.rmtree(path + "/gif_images")
- get_accuracy_graph(self.voter_perfs, self.__class__.__name__,
- directory + 'voter_perfs.png', "Errors")
+ # try:
+ # import os
+ # os.makedirs(path + "/gif_images")
+ # except:
+ # raise
+ # filenames = []
+ # max_weight = max([np.max(examples_weights) for examples_weights in
+ # self.example_weights_])
+ # min_weight = min([np.max(examples_weights) for examples_weights in
+ # self.example_weights_])
+ # for iterIndex, examples_weights in enumerate(self.example_weights_):
+ # r = np.array(examples_weights)
+ # theta = np.arange(self.n_total_examples)
+ # colors = np.sign(self.previous_margins[iterIndex])
+ # fig = plt.figure(figsize=(5, 5), dpi=80)
+ # ax = fig.add_subplot(111)
+ # c = ax.scatter(theta, r, c=colors, cmap='RdYlGn', alpha=0.75)
+ # ax.set_ylim(min_weight, max_weight)
+ # filename = path + "/gif_images/" + str(iterIndex) + ".png"
+ # filenames.append(filename)
+ # plt.savefig(filename)
+ # plt.close()
+ #
+ # import imageio
+ # images = []
+ # logging.getLogger("PIL").setLevel(logging.WARNING)
+ # for filename in filenames:
+ # images.append(imageio.imread(filename))
+ # imageio.mimsave(path + '/weights.gif', images, duration=1. / 2)
+ # import shutil
+ # shutil.rmtree(path + "/gif_images")
+ get_accuracy_graph(self.voter_perfs[:20], self.__class__.__name__,
+ directory + 'voter_perfs.png', "Rs")
+ get_accuracy_graph(self.weights_, self.__class__.__name__,
+ directory+'vote_weights.png', "weights")
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")
+ directory + 'margins.png', "Squared Margins")
+ print(self.selected_margins)
+ print(len(self.selected_margins))
+ get_accuracy_graph(self.selected_margins, self.__class__.__name__,
+ directory + 'selected_margins.png', "Selected 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:])
+ 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:
diff --git a/multiview_platform/MonoMultiViewClassifiers/Monoview/ExecClassifMonoView.py b/multiview_platform/MonoMultiViewClassifiers/Monoview/ExecClassifMonoView.py
index fd0784a91ad017bad6844b79b560212b1d614149..95f89ec7f8b9cb2787e6f2f83833f9b7a13545d7 100644
--- a/multiview_platform/MonoMultiViewClassifiers/Monoview/ExecClassifMonoView.py
+++ b/multiview_platform/MonoMultiViewClassifiers/Monoview/ExecClassifMonoView.py
@@ -106,7 +106,7 @@ def ExecMonoview(directory, X, Y, name, labelsNames, classificationIndices, KFol
logging.debug("Done:\t Getting Results")
logging.debug("Start:\t Saving preds")
- saveResults(stringAnalysis, outputFileName, full_labels_pred, y_train_pred, y_train, imagesAnalysis)
+ saveResults(stringAnalysis, outputFileName, full_labels_pred, y_train_pred, y_train, imagesAnalysis, y_test)
logging.info("Done:\t Saving Results")
viewIndex = args["viewIndex"]
@@ -172,7 +172,7 @@ def getHPs(classifierModule, hyperParamSearch, nIter, CL_type, X_train, y_train,
return clKWARGS, testFoldsPreds
-def saveResults(stringAnalysis, outputFileName, full_labels_pred, y_train_pred, y_train, imagesAnalysis):
+def saveResults(stringAnalysis, outputFileName, full_labels_pred, y_train_pred, y_train, imagesAnalysis, y_test):
logging.info(stringAnalysis)
outputTextFile = open(outputFileName + 'summary.txt', 'w')
outputTextFile.write(stringAnalysis)
@@ -180,6 +180,8 @@ def saveResults(stringAnalysis, outputFileName, full_labels_pred, y_train_pred,
np.savetxt(outputFileName + "full_pred.csv", full_labels_pred.astype(np.int16), delimiter=",")
np.savetxt(outputFileName + "train_pred.csv", y_train_pred.astype(np.int16), delimiter=",")
np.savetxt(outputFileName + "train_labels.csv", y_train.astype(np.int16), delimiter=",")
+ np.savetxt(outputFileName + "test_labels.csv", y_test.astype(np.int16),
+ delimiter=",")
if imagesAnalysis is not None:
for imageName in imagesAnalysis:
diff --git a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC.py b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC.py
index 6bf35c91ce62e5a77589e1484715f5b125a03f07..2ea8b44664323f7656ab1840774f31a4a5b77a65 100644
--- a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC.py
+++ b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC.py
@@ -6,15 +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=500,
+ super(QarBoostNC, self).__init__(n_max_iterations=300,
random_state=random_state,
self_complemented=True,
- twice_the_same=False,
+ twice_the_same=True,
c_bound_choice=True,
random_start=False,
n_stumps_per_attribute=1,
use_r=True,
- c_bound_sol=False
+ c_bound_sol=True
)
self.param_names = []
self.distribs = []
diff --git a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC2.py b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC2.py
index e27a8a4d045f9c11a19350e00860e3f8d021fb75..9b07ade50a40f220a210fbfa1bc047002dc97c30 100644
--- a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC2.py
+++ b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC2.py
@@ -6,7 +6,7 @@ from ..Monoview.Additions.QarBoostUtils import ColumnGenerationClassifierQar
class QarBoostNC2(ColumnGenerationClassifierQar, BaseMonoviewClassifier):
def __init__(self, random_state=None, **kwargs):
- super(QarBoostNC2, self).__init__(n_max_iterations=500,
+ super(QarBoostNC2, self).__init__(n_max_iterations=300,
random_state=random_state,
self_complemented=True,
twice_the_same=True,
diff --git a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC3.py b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC3.py
index 1c44cc3b6414e97269b31d71b125a5d0c82d8d6b..7133f80620bbae532461b5f7b053d3a8b440157a 100644
--- a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC3.py
+++ b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostNC3.py
@@ -6,7 +6,7 @@ from ..Monoview.Additions.QarBoostUtils import ColumnGenerationClassifierQar
class QarBoostNC3(ColumnGenerationClassifierQar, BaseMonoviewClassifier):
def __init__(self, random_state=None, **kwargs):
- super(QarBoostNC3, self).__init__(n_max_iterations=500,
+ super(QarBoostNC3, self).__init__(n_max_iterations=300,
random_state=random_state,
self_complemented=True,
twice_the_same=False,
@@ -14,7 +14,8 @@ class QarBoostNC3(ColumnGenerationClassifierQar, BaseMonoviewClassifier):
random_start=False,
n_stumps_per_attribute=1,
use_r=True,
- c_bound_sol=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 e06e069841c186eaab2cee6ff090cdc42f6b2fa3..934e07ad329bf0f6e11546caab967e71163eb798 100644
--- a/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostv2.py
+++ b/multiview_platform/MonoMultiViewClassifiers/MonoviewClassifiers/QarBoostv2.py
@@ -6,15 +6,15 @@ from ..Monoview.Additions.QarBoostUtils import ColumnGenerationClassifierQar
class QarBoostv2(ColumnGenerationClassifierQar, BaseMonoviewClassifier):
def __init__(self, random_state=None, **kwargs):
- super(QarBoostv2, self).__init__(n_max_iterations=500,
+ super(QarBoostv2, self).__init__(n_max_iterations=300,
random_state=random_state,
self_complemented=True,
- twice_the_same=True,
+ twice_the_same=False,
c_bound_choice=True,
random_start=False,
n_stumps_per_attribute=1,
use_r=True,
- c_bound_sol=True
+ c_bound_sol=False
)
self.param_names = []
self.distribs = []
diff --git a/multiview_platform/MonoMultiViewClassifiers/utils/Dataset.py b/multiview_platform/MonoMultiViewClassifiers/utils/Dataset.py
index 686cdc199aa3db43193d109ef51d3eae7d4a280a..f11b060f303bf357ce5e18a467466fd509796f84 100644
--- a/multiview_platform/MonoMultiViewClassifiers/utils/Dataset.py
+++ b/multiview_platform/MonoMultiViewClassifiers/utils/Dataset.py
@@ -40,6 +40,7 @@ def getShape(DATASET, viewIndex):
return DATASET.get("View" + str(viewIndex)).attrs["shape"]
+
def getValue(DATASET):
"""Used to get the value of a view in the HDF5 dataset even if it sparse"""
if not DATASET.attrs["sparse"]: