From acc0dd2358831f698dd26610710f18559aa74a70 Mon Sep 17 00:00:00 2001
From: Baptiste Bauvin <baptiste.bauvin@lis-lab.fr>
Date: Wed, 22 Jun 2022 07:37:52 -0400
Subject: [PATCH] Cuisine
---
summit/multiview_platform/exec_classif.py | 6 ++--
.../metrics/specificity_score.py | 23 +++++++++++++
.../monoview/exec_classif_mono_view.py | 8 +++++
.../monoview_classifiers/imbalance_bagging.py | 12 ++++++-
.../monoview_classifiers/samba.py | 15 ++++++---
.../monoview_classifiers/scm_bagging_mincq.py | 6 ++--
.../result_analysis/feature_importances.py | 3 +-
summit/multiview_platform/utils/base.py | 32 +++++++++++--------
summit/multiview_platform/utils/dataset.py | 14 +++++---
9 files changed, 89 insertions(+), 30 deletions(-)
create mode 100644 summit/multiview_platform/metrics/specificity_score.py
diff --git a/summit/multiview_platform/exec_classif.py b/summit/multiview_platform/exec_classif.py
index f742a7fd..11697f4b 100644
--- a/summit/multiview_platform/exec_classif.py
+++ b/summit/multiview_platform/exec_classif.py
@@ -640,9 +640,9 @@ def exec_classif(arguments): # pragma: no cover
k_folds = execution.gen_k_folds(stats_iter, args["nb_folds"],
stats_iter_random_states)
- dataset_files = dataset.init_multiple_datasets(args["pathf"],
- args["name"],
- nb_cores)
+ # dataset_files = dataset.init_multiple_datasets(args["pathf"],
+ # args["name"],
+ # nb_cores)
views, views_indices, all_views = execution.init_views(dataset_var,
args[
diff --git a/summit/multiview_platform/metrics/specificity_score.py b/summit/multiview_platform/metrics/specificity_score.py
new file mode 100644
index 00000000..d9f6f585
--- /dev/null
+++ b/summit/multiview_platform/metrics/specificity_score.py
@@ -0,0 +1,23 @@
+from sklearn.metrics import make_scorer
+from sklearn.metrics import confusion_matrix as metric
+
+# Author-Info
+__author__ = "Baptiste Bauvin"
+__status__ = "Prototype" # Production, Development, Prototype
+
+
+def score(y_true, y_pred, **kwargs):
+ score = metric(y_true, y_pred, **kwargs)
+ if score[0,0]+score[0,1] !=0:
+ return score[0,0]/(score[0,0]+score[0,1])
+ else:
+ return 0
+
+
+def get_scorer(**kwargs):
+ return make_scorer(score, greater_is_better=True, **kwargs)
+
+
+def get_config(**kwargs):
+ configString = "Specificity score (higher is better)".format(kwargs)
+ return configString
\ No newline at end of file
diff --git a/summit/multiview_platform/monoview/exec_classif_mono_view.py b/summit/multiview_platform/monoview/exec_classif_mono_view.py
index 33b5a32c..44ef75b2 100644
--- a/summit/multiview_platform/monoview/exec_classif_mono_view.py
+++ b/summit/multiview_platform/monoview/exec_classif_mono_view.py
@@ -120,6 +120,14 @@ def exec_monoview(directory, X, Y, database_name, labels_names,
test_pred = classifier.predict(X_test)
pred_duration = time.monotonic() - pred_beg
+ #### ROC CURVE ADDITION ###
+ from sklearn.metrics import roc_curve
+ fpr, tpr, _ = roc_curve(y_test, classifier.predict_proba(X_test)[:, 1])
+ np.savetxt(os.path.join(directory, classifier_class_name+"-fpr.npy"), fpr)
+ np.savetxt(os.path.join(directory, classifier_class_name + "-tpr.npy"), tpr)
+ ### END ROC ###
+
+
# Filling the full prediction in the right order
full_pred = np.zeros(Y.shape, dtype=int) - 100
for train_index, index in enumerate(classification_indices[0]):
diff --git a/summit/multiview_platform/monoview_classifiers/imbalance_bagging.py b/summit/multiview_platform/monoview_classifiers/imbalance_bagging.py
index c4340420..9dfa2e26 100644
--- a/summit/multiview_platform/monoview_classifiers/imbalance_bagging.py
+++ b/summit/multiview_platform/monoview_classifiers/imbalance_bagging.py
@@ -1,5 +1,6 @@
from imblearn.ensemble import BalancedBaggingClassifier
-from sklearn.tree import DecisionTreeClassifier
+import numpy as np
+
from ..monoview.monoview_utils import BaseMonoviewClassifier
from ..utils.base import base_boosting_estimators
@@ -27,5 +28,14 @@ class ImbalanceBagging(BaseMonoviewClassifier, BalancedBaggingClassifier):
self.weird_strings = {"base_estimator": "class_name"}
self.base_estimator_config = base_estimator_config
+ def fit(self, X, y):
+ BalancedBaggingClassifier.fit(self, X, y)
+ self.feature_importances_ = np.zeros(X.shape[1])
+ for estim in self.estimators_:
+ if hasattr(estim['classifier'], 'feature_importances_'):
+ self.feature_importances_ += estim['classifier'].feature_importances_
+ self.feature_importances_ /= np.sum(self.feature_importances_)
+ return self
+
diff --git a/summit/multiview_platform/monoview_classifiers/samba.py b/summit/multiview_platform/monoview_classifiers/samba.py
index 97094c6d..a69de937 100644
--- a/summit/multiview_platform/monoview_classifiers/samba.py
+++ b/summit/multiview_platform/monoview_classifiers/samba.py
@@ -33,6 +33,7 @@ class SamBAClf(NeighborHoodClassifier, BaseMonoviewClassifier):
pred_train=False,
forced_diversity=False,
normalize_dists=False,
+ class_weight="balanced",
**kwargs):
"""
@@ -54,18 +55,22 @@ class SamBAClf(NeighborHoodClassifier, BaseMonoviewClassifier):
normalizer=normalizer,
forced_diversity=forced_diversity,
b=b, a=a, pred_train=pred_train,
- normalize_dists=normalize_dists)
- self.param_names = ["n_estimators", "relevance", "distance",
+ normalize_dists=normalize_dists,
+ class_weight=class_weight)
+ self.param_names = ["n_estimators",
+ "relevance",
+ "distance",
"train_weighting", "b", "pred_train", "normalizer",
- "normalize_dists", "a"]
+ "normalize_dists", "a", "class_weight"]
self.distribs = [CustomRandint(low=1, high=70),
[ExpRelevance()],
- [EuclidianDist(), PolarDist(), ExpEuclidianDist()],
+ [EuclidianDist(), PolarDist(), ExpEuclidianDist(), Jaccard()],
[ExpTrainWeighting()],
CustomUniform(0.1, 6,),
[True, False],
[RobustScaler()],
- [True], CustomRandint(0, 10, 'e-')]
+ [True], CustomRandint(0, 10, 'e-'),
+ ["balanced", None]]
self.classed_params = []
self.weird_strings = {}
diff --git a/summit/multiview_platform/monoview_classifiers/scm_bagging_mincq.py b/summit/multiview_platform/monoview_classifiers/scm_bagging_mincq.py
index b4b38d60..32892dd1 100644
--- a/summit/multiview_platform/monoview_classifiers/scm_bagging_mincq.py
+++ b/summit/multiview_platform/monoview_classifiers/scm_bagging_mincq.py
@@ -1,4 +1,5 @@
-from scm_bagging.scm_bagging_classifier import ScmBaggingClassifier
+from randomscm.randomscm import RandomScmClassifier
+
from ..monoview.monoview_utils import BaseMonoviewClassifier
@@ -16,7 +17,8 @@ from six import iteritems
MAX_INT = np.iinfo(np.int32).max
-class ScmBaggingMinCq(ScmBaggingClassifier, BaseMonoviewClassifier):
+class ScmBaggingMinCq(RandomScmClassifier, BaseMonoviewClassifier):
+
"""A Bagging classifier. for SetCoveringMachineClassifier()
The base estimators are built on subsets of both samples
and features.
diff --git a/summit/multiview_platform/result_analysis/feature_importances.py b/summit/multiview_platform/result_analysis/feature_importances.py
index 36c0eb35..042e4c0d 100644
--- a/summit/multiview_platform/result_analysis/feature_importances.py
+++ b/summit/multiview_platform/result_analysis/feature_importances.py
@@ -140,7 +140,8 @@ def plot_feature_relevance(file_name, feature_importance,
for score in score_df.columns:
if len(score.split("-"))>1:
algo, view = score.split("-")
- feature_importance[algo].loc[[ind for ind in feature_importance.index if ind.startswith(view)]]*=score_df[score]['test']
+ list_ind = [ind for ind in feature_importance.index if ind.startswith(view)]
+ feature_importance[algo].loc[list_ind]*=2*(score_df[score]['test']-0.5)
else:
feature_importance[score] *= score_df[score]['test']
file_name+="_relevance"
diff --git a/summit/multiview_platform/utils/base.py b/summit/multiview_platform/utils/base.py
index 67df47a6..a32afbd0 100644
--- a/summit/multiview_platform/utils/base.py
+++ b/summit/multiview_platform/utils/base.py
@@ -253,21 +253,27 @@ class ResultAnalyser():
metric_module = getattr(metrics, metric)
else:
metric_module = getattr(metrics, metric[:-1])
+
class_train_scores = []
class_test_scores = []
- for label_value in np.unique(self.labels):
- train_sample_indices = self.train_indices[
- np.where(self.labels[self.train_indices] == label_value)[0]]
- test_sample_indices = self.test_indices[
- np.where(self.labels[self.test_indices] == label_value)[0]]
- class_train_scores.append(
- metric_module.score(y_true=self.labels[train_sample_indices],
- y_pred=self.pred[train_sample_indices],
- **metric_kwargs))
- class_test_scores.append(
- metric_module.score(y_true=self.labels[test_sample_indices],
- y_pred=self.pred[test_sample_indices],
- **metric_kwargs))
+ if metric not in ["roc_auc_score", "specificity_score"]:
+ for label_value in np.unique(self.labels):
+ train_sample_indices = self.train_indices[
+ np.where(self.labels[self.train_indices] == label_value)[0]]
+ test_sample_indices = self.test_indices[
+ np.where(self.labels[self.test_indices] == label_value)[0]]
+ class_train_scores.append(
+ metric_module.score(y_true=self.labels[train_sample_indices],
+ y_pred=self.pred[train_sample_indices],
+ **metric_kwargs))
+ class_test_scores.append(
+ metric_module.score(y_true=self.labels[test_sample_indices],
+ y_pred=self.pred[test_sample_indices],
+ **metric_kwargs))
+ else:
+ for _ in np.unique(self.labels):
+ class_train_scores.append(0)
+ class_test_scores.append(0)
train_score = metric_module.score(
y_true=self.labels[self.train_indices],
y_pred=self.pred[self.train_indices],
diff --git a/summit/multiview_platform/utils/dataset.py b/summit/multiview_platform/utils/dataset.py
index 15175976..600a0661 100644
--- a/summit/multiview_platform/utils/dataset.py
+++ b/summit/multiview_platform/utils/dataset.py
@@ -458,11 +458,11 @@ class HDF5Dataset(Dataset):
for view_index in range(self.nb_view):
if "feature_ids-View{}".format(view_index) in self.dataset["Metadata"].keys():
self.feature_ids[view_index] = [feature_id.decode()
- if not is_just_number(feature_id.decode())
- else "ID_" + feature_id.decode()
- for feature_id in self.dataset["Metadata"]["feature_ids-View{}".format(view_index)]]
+ if not is_just_number(feature_id.decode())
+ else "ID_" + feature_id.decode()
+ for feature_id in self.dataset["Metadata"]["feature_ids-View{}".format(view_index)]]
else:
- self.gen_feat_id(view_index)
+ self.gen_feat_id(view_index)
def get_nb_samples(self):
"""
@@ -503,7 +503,7 @@ class HDF5Dataset(Dataset):
seleted labels' names
"""
selected_labels = self.get_labels(sample_indices)
- if decode:
+ if type(self.dataset["Labels"].attrs["names"][0]) == bytes:
return [label_name.decode("utf-8")
for label, label_name in
enumerate(self.dataset["Labels"].attrs["names"])
@@ -619,10 +619,14 @@ class HDF5Dataset(Dataset):
view_names = self.init_view_names(view_names)
new_dataset_file["Metadata"].attrs["nbView"] = len(view_names)
for new_index, view_name in enumerate(view_names):
+ del new_dataset_file["Metadata"]["feature_ids-View{}".format(new_index)]
+ new_dataset_file["Metadata"]["feature_ids-View{}".format(new_index)] = new_dataset_file["Metadata"]["feature_ids-View{}".format(self.view_dict[view_name])]
+ del new_dataset_file["Metadata"]["feature_ids-View{}".format(self.view_dict[view_name])]
self.copy_view(target_dataset=new_dataset_file,
source_view_name=view_name,
target_view_index=new_index,
sample_indices=sample_indices)
+
new_dataset_file.close()
self.update_hdf5_dataset(dataset_file_path)
--
GitLab