From e5c042815ab517750d438a66798abe2e268e1e5e Mon Sep 17 00:00:00 2001
From: Baptiste Bauvin <baptiste.bauvin@lis-lab.fr>
Date: Thu, 28 Oct 2021 07:57:48 -0400
Subject: [PATCH] added some files
---
config_files/config_cuisine.yml | 88 ++++++++++++++-----
config_files/config_private_algos.yml | 26 +++---
setup.py | 1 +
summit/multiview_platform/exec_classif.py | 6 +-
.../metrics/balanced_accuracy.py | 41 +++++++++
.../monoview_classifiers/adaboost.py | 2 +-
.../monoview_classifiers/adaboost_pregen.py | 1 +
.../monoview_classifiers/samba.py | 70 +++++++++++++++
.../early_fusion_samba.py | 40 +++++++++
.../multiview_classifiers/mumbo.py | 7 ++
10 files changed, 246 insertions(+), 36 deletions(-)
create mode 100644 summit/multiview_platform/metrics/balanced_accuracy.py
create mode 100644 summit/multiview_platform/monoview_classifiers/samba.py
create mode 100644 summit/multiview_platform/multiview_classifiers/early_fusion_samba.py
diff --git a/config_files/config_cuisine.yml b/config_files/config_cuisine.yml
index 5a916146..9206cf2f 100644
--- a/config_files/config_cuisine.yml
+++ b/config_files/config_cuisine.yml
@@ -1,10 +1,10 @@
# The base configuration of the benchmark
log: True
-name: ["ionosphere", "abalone", "australian", "balance", "bupa", "cylinder", "hepatitis", "pima", "yeast", "zoo"]
-label: "comp_1"
+name: ['tnbc_mazid']
+label: ""
file_type: ".hdf5"
views:
-pathf: "/home/baptiste/Documents/Datasets/UCI/both/"
+pathf: "/home/baptiste/Documents/Datasets/Mazid/"
nice: 0
random_state: 42
nb_cores: 1
@@ -13,34 +13,53 @@ debug: True
add_noise: False
noise_std: 0.0
res_dir: "../results/"
-track_tracebacks: False
+track_tracebacks: True
# All the classification-realted configuration options
multiclass_method: "oneVersusOne"
-split: 0.50
+split: 0.30
nb_folds: 5
nb_class: 2
classes:
-type: ["monoview",]
-algos_monoview: ["cb_boost", "self_opt_cb", "adaboost", "cq_boost", "min_cq", "adaboost_pregen", "self_opt_cb_pseudo", "self_opt_cb_root"]
-algos_multiview: ["mv_cb_boost","early_fusion_dt", "early_fusion_cb", "early_fusion_rf","mumbo", "early_fusion_svm" ]
+type: ["monoview","multiview"]
+algos_monoview: ["samba", "scm_bagging", "random_forest", "adaboost", 'scm']
+algos_multiview: ["early_fusion_adaboost", "early_fusion_decision_tree", "early_fusion_random_forest", "early_fusion_samba"]
stats_iter: 5
metrics:
- accuracy_score: {}
+ balanced_accuracy: {}
f1_score:
average: 'micro'
-metric_princ: "accuracy_score"
-hps_type: "None"
+ accuracy_score: {}
+metric_princ: "balanced_accuracy"
+hps_type: "Random"
hps_args:
- n_iter: 30
- equivalent_draws: True
+ n_iter: 20
+ equivalent_draws: False
svm_rbf:
C: 0.7
+scm_bagging:
+ {max_features: 0.908115713423863, max_rules: 9, max_samples: 0.9277949143533335, model_type: conjunction,
+ n_estimators: 109, p_options: 0.7823433255515356}
+
+samba:
+ n_estimators: 22
+
+adaboost:
+ {base_estimator: DecisionTreeClassifier, base_estimator__ccp_alpha: 0.0, base_estimator__class_weight: null,
+ base_estimator__criterion: gini, base_estimator__max_depth: 5, base_estimator__max_features: null,
+ base_estimator__max_leaf_nodes: null, base_estimator__min_impurity_decrease: 0.0,
+ base_estimator__min_impurity_split: null, base_estimator__min_samples_leaf: 1, base_estimator__min_samples_split: 2,
+ base_estimator__min_weight_fraction_leaf: 0.0, base_estimator__random_state: null,
+ base_estimator__splitter: best, n_estimators: 354}
+
+svm_linear:
+ C: 0.3867
+
cb_boost:
n_stumps: 1
- n_max_iterations: 10
+ n_max_iterations: 20
estimators_generator: "Stumps"
cq_boost:
@@ -50,15 +69,42 @@ cq_boost:
min_cq:
n_stumps_per_attribute: 1
-adaboost:
- n_estimators: 10
+decision_tree:
+ {criterion: entropy, max_depth: 271, splitter: random}
-adaboost_pregen:
- n_estimators: 10
- n_stumps: 1
+early_fusion_adaboost:
+ {base_estimator: DecisionTreeClassifier, base_estimator__ccp_alpha: 0.0, base_estimator__class_weight: null,
+ base_estimator__criterion: gini, base_estimator__max_depth: 5, base_estimator__max_features: null,
+ base_estimator__max_leaf_nodes: null, base_estimator__min_impurity_decrease: 0.0,
+ base_estimator__min_impurity_split: null, base_estimator__min_samples_leaf: 1, base_estimator__min_samples_split: 2,
+ base_estimator__min_weight_fraction_leaf: 0.0, base_estimator__random_state: null,
+ base_estimator__splitter: best, base_estimator_config: null, n_estimators: 273}
-decision_tree:
- max_depth: 2
+early_fusion_decision_tree:
+ {criterion: entropy, max_depth: 293, splitter: random}
+
+early_fusion_random_forest:
+ {criterion: gini, max_depth: 8, n_estimators: 46}
+
+random_forest:
+ {criterion: gini, max_depth: 8, n_estimators: 32}
+
+weighted_linear_late_fusion:
+ classifier_configs:
+ - decision_tree: {criterion: entropy, max_depth: 112, splitter: random}
+ - adaboost: {base_estimator: DecisionTreeClassifier, base_estimator__ccp_alpha: 0.0,
+ base_estimator__class_weight: null, base_estimator__criterion: gini, base_estimator__max_depth: 2,
+ base_estimator__max_features: null, base_estimator__max_leaf_nodes: null, base_estimator__min_impurity_decrease: 0.0,
+ base_estimator__min_impurity_split: null, base_estimator__min_samples_leaf: 1,
+ base_estimator__min_samples_split: 2, base_estimator__min_weight_fraction_leaf: 0.0,
+ base_estimator__random_state: null, base_estimator__splitter: best, n_estimators: 400}
+ classifiers_names: [decision_tree, adaboost]
+ nb_cores: 1
+ rs: 724
+ weights: [0.9636627605010293, 0.3834415188257777]
+
+scm:
+ {max_rules: 10, model_type: conjunction, p: 0.8310271995093625}
mumbo:
base_estimator:
diff --git a/config_files/config_private_algos.yml b/config_files/config_private_algos.yml
index bc745606..a676749a 100644
--- a/config_files/config_private_algos.yml
+++ b/config_files/config_private_algos.yml
@@ -1,14 +1,14 @@
# The base configuration of the benchmark
log: True
-name: ["mnist_0_9_train"]
+name: ["multiview_mnist"]
label: "_"
file_type: ".hdf5"
-views: ["NIMST_data", ]
-pathf: "/home/baptiste/Documents/Datasets/MNist/"
+views:
+pathf: "examples/data/"
nice: 0
random_state: 43
nb_cores: 1
-full: False
+full: True
debug: True
add_noise: False
noise_std: 0.0
@@ -19,16 +19,20 @@ track_tracebacks: False
multiclass_method: "oneVersusOne"
split: 0.96
nb_folds: 5
-nb_class: 2
+nb_class:
classes:
-type: ["monoview",]
-algos_monoview: ["hm_gb_cbound","cb_boost"]
-algos_multiview: ["mumbo","mvml"]
+type: ["monoview","multiview"]
+algos_monoview: ["decision_tree","adaboost"]
+algos_multiview: ["mumbo","mvml", 'lp_norm_mkl', 'mucombo', 'early_fusion_decision_tree', 'early_fusion_adaboost']
stats_iter: 1
metrics:
- zero_one_loss: {}
+ accuracy_score: {}
f1_score: {}
-metric_princ: "zero_one_loss"
+metric_princ: "accuracy_score"
hps_type: "None"
hps_args:
- n_iter: 2
\ No newline at end of file
+ n_iter: 2
+mumbo:
+ base_estimator:
+ decision_tree:
+ max_depth: 3
\ No newline at end of file
diff --git a/setup.py b/setup.py
index c92d5a75..8f774ddf 100644
--- a/setup.py
+++ b/setup.py
@@ -94,6 +94,7 @@ def setup_package():
# ce qui est notre cas
license="GNUGPL",
+
# Il y a encore une chiée de paramètres possibles, mais avec ça vous
# couvrez 90% des besoins
# ext_modules=cythonize(
diff --git a/summit/multiview_platform/exec_classif.py b/summit/multiview_platform/exec_classif.py
index 6c75194a..a84014d3 100644
--- a/summit/multiview_platform/exec_classif.py
+++ b/summit/multiview_platform/exec_classif.py
@@ -548,7 +548,7 @@ def exec_one_benchmark_mono_core(dataset_var=None, labels_dictionary=None,
argument_dictionaries=None,
benchmark=None, views=None, views_indices=None,
flag=None, labels=None,
- track_tracebacks=False): # pragma: no cover
+ track_tracebacks=False, nb_cores=1): # pragma: no cover
results_monoview, labels_names = benchmark_init(directory,
classification_indices,
labels,
@@ -564,7 +564,7 @@ def exec_one_benchmark_mono_core(dataset_var=None, labels_dictionary=None,
results_monoview += [
exec_monoview(directory, X, Y, args["name"], labels_names,
classification_indices, k_folds,
- 1, args["file_type"], args["pathf"], random_state,
+ nb_cores, args["file_type"], args["pathf"], random_state,
hyper_param_search=hyper_param_search,
metrics=metrics,
**arguments)]
@@ -679,7 +679,7 @@ def exec_benchmark(nb_cores, stats_iter,
for arguments in benchmark_arguments_dictionaries:
benchmark_results = exec_one_benchmark_mono_core(
dataset_var=dataset_var,
- track_tracebacks=track_tracebacks,
+ track_tracebacks=track_tracebacks, nb_cores=nb_cores,
**arguments)
analyze_iterations([benchmark_results],
benchmark_arguments_dictionaries, stats_iter,
diff --git a/summit/multiview_platform/metrics/balanced_accuracy.py b/summit/multiview_platform/metrics/balanced_accuracy.py
new file mode 100644
index 00000000..6d4ab5d0
--- /dev/null
+++ b/summit/multiview_platform/metrics/balanced_accuracy.py
@@ -0,0 +1,41 @@
+"""Functions :
+ score: to get the accuracy score
+ get_scorer: returns a sklearn scorer for grid search
+"""
+
+from sklearn.metrics import balanced_accuracy_score as metric
+from sklearn.metrics import make_scorer
+
+# Author-Info
+__author__ = "Baptiste Bauvin"
+__status__ = "Prototype" # Production, Development, Prototype
+
+
+def score(y_true, y_pred, multiclass=False, **kwargs):
+ """Arguments:
+ y_true: real labels
+ y_pred: predicted labels
+
+ Keyword Arguments:
+ "0": weights to compute accuracy
+
+ Returns:
+ Weighted accuracy score for y_true, y_pred"""
+ score = metric(y_true, y_pred, **kwargs)
+ return score
+
+
+def get_scorer(**kwargs):
+ """Keyword Arguments:
+ "0": weights to compute accuracy
+
+ Returns:
+ A weighted sklearn scorer for accuracy"""
+ return make_scorer(metric, greater_is_better=True,
+ **kwargs)
+
+
+def get_config(**kwargs):
+ config_string = "Balanced accuracy score using {}, (higher is better)".format(
+ kwargs)
+ return config_string
diff --git a/summit/multiview_platform/monoview_classifiers/adaboost.py b/summit/multiview_platform/monoview_classifiers/adaboost.py
index cd8ce3db..579b9ffd 100644
--- a/summit/multiview_platform/monoview_classifiers/adaboost.py
+++ b/summit/multiview_platform/monoview_classifiers/adaboost.py
@@ -35,7 +35,7 @@ class Adaboost(AdaBoostClassifier, BaseMonoviewClassifier):
)
self.param_names = ["n_estimators", "base_estimator"]
self.classed_params = ["base_estimator"]
- self.distribs = [CustomRandint(low=1, high=500),
+ self.distribs = [CustomRandint(low=1, high=100),
base_boosting_estimators]
self.weird_strings = {"base_estimator": "class_name"}
self.plotted_metric = metrics.zero_one_loss
diff --git a/summit/multiview_platform/monoview_classifiers/adaboost_pregen.py b/summit/multiview_platform/monoview_classifiers/adaboost_pregen.py
index f0fbd955..43589981 100644
--- a/summit/multiview_platform/monoview_classifiers/adaboost_pregen.py
+++ b/summit/multiview_platform/monoview_classifiers/adaboost_pregen.py
@@ -17,6 +17,7 @@ __status__ = "Prototype" # Production, Development, Prototype
classifier_class_name = "AdaboostPregen"
+
class AdaboostPregen(AdaBoostClassifier, BaseMonoviewClassifier,
PregenClassifier):
"""
diff --git a/summit/multiview_platform/monoview_classifiers/samba.py b/summit/multiview_platform/monoview_classifiers/samba.py
new file mode 100644
index 00000000..f43defd7
--- /dev/null
+++ b/summit/multiview_platform/monoview_classifiers/samba.py
@@ -0,0 +1,70 @@
+from SamBA.samba import NeighborHoodClassifier, ExpTrainWeighting
+import numpy as np
+from sklearn.tree import DecisionTreeClassifier
+from SamBA.relevances import *
+from SamBA.distances import *
+from sklearn.preprocessing import RobustScaler
+
+from ..monoview.monoview_utils import BaseMonoviewClassifier
+from ..utils.hyper_parameter_search import CustomRandint, CustomUniform
+
+
+# Author-Info
+__author__ = "Baptiste Bauvin"
+__status__ = "Prototype" # Production, Development, Prototype
+
+
+# class Decis
+classifier_class_name = "SamBAClf"
+
+class SamBAClf(NeighborHoodClassifier, BaseMonoviewClassifier):
+
+ def __init__(self, base_estimator=DecisionTreeClassifier(max_depth=1,
+ splitter='best',
+ criterion='gini'),
+ n_estimators=2,
+ estimator_params=tuple(),
+ relevance=MarginRelevance(),
+ distance=EuclidianDist(),
+ train_weighting=ExpTrainWeighting(),
+ keep_selected_features=True,
+ normalizer=RobustScaler(),
+ b=2,
+ pred_train=False,
+ forced_diversity=True,
+ **kwargs):
+ """
+
+ Parameters
+ ----------
+ random_state
+ model_type
+ max_rules
+ p
+ kwargs
+ """
+ super(SamBAClf, self).__init__(base_estimator=base_estimator,
+ n_estimators=n_estimators,
+ estimator_params=estimator_params,
+ relevance=relevance,
+ distance=distance,
+ train_weighting=train_weighting,
+ keep_selected_features=keep_selected_features,
+ normalizer=normalizer,
+ forced_diversity=forced_diversity,
+ b=b,pred_train=pred_train)
+ self.param_names = ["n_estimators", "relevance", "distance",
+ "train_weighting", "b", "pred_train", "normalizer"]
+ self.distribs = [CustomRandint(low=1, high=30),
+ [ExpRelevance(), MarginRelevance()],
+ [EuclidianDist(), PolarDist(), ExpEuclidianDist()],
+ [ExpTrainWeighting()],
+ CustomUniform(0.5, 3),
+ [True, False],
+ [RobustScaler(), None]]
+ self.classed_params = []
+ self.weird_strings = {}
+
+ def get_interpretation(self, directory, base_file_name, y_test, multi_class=False):
+ interpret_string = self.get_feature_importance(directory, base_file_name)
+ return interpret_string
diff --git a/summit/multiview_platform/multiview_classifiers/early_fusion_samba.py b/summit/multiview_platform/multiview_classifiers/early_fusion_samba.py
new file mode 100644
index 00000000..7722f003
--- /dev/null
+++ b/summit/multiview_platform/multiview_classifiers/early_fusion_samba.py
@@ -0,0 +1,40 @@
+from .additions.early_fusion_from_monoview import BaseEarlyFusion
+from ..utils.hyper_parameter_search import CustomRandint, CustomUniform
+from ..monoview_classifiers.samba import SamBAClf
+from SamBA.samba import *
+from SamBA.distances import *
+from SamBA.relevances import *
+from sklearn.tree import DecisionTreeClassifier
+
+classifier_class_name = "EarlyFusionSamba"
+
+
+class EarlyFusionSamba(BaseEarlyFusion):
+
+ def __init__(self, random_state=None, base_estimator=DecisionTreeClassifier(max_depth=1,
+ splitter='best',
+ criterion='gini'),
+ n_estimators=2,
+ estimator_params=tuple(),
+ relevance=MarginRelevance(),
+ distance=EuclidianDist(),
+ train_weighting=ExpTrainWeighting(pred_train=True),
+ keep_selected_features=True,
+ normalizer=RobustScaler(),
+ pred_train=False,
+ b=2,
+ **kwargs):
+ BaseEarlyFusion.__init__(self, random_state=random_state,
+ monoview_classifier="samba",
+ base_estimator=base_estimator, estimator_params=estimator_params,
+ relevance=relevance, distance=distance, train_weighting=train_weighting,
+ keep_selected_features=keep_selected_features, normalizer=normalizer,
+ n_estimators=n_estimators, pred_train=pred_train, b=b, **kwargs)
+ self.param_names = ["n_estimators", "relevance", "distance",
+ "train_weighting", "b", "pred_train"]
+ self.distribs = [CustomRandint(low=1, high=30),
+ [ExpRelevance(), MarginRelevance()],
+ [EuclidianDist(), PolarDist(), ExpEuclidianDist()],
+ [ExpTrainWeighting(pred_train=True)],
+ CustomUniform(0.25, 3),
+ [True, False]]
\ No newline at end of file
diff --git a/summit/multiview_platform/multiview_classifiers/mumbo.py b/summit/multiview_platform/multiview_classifiers/mumbo.py
index e631cbc2..203228ba 100644
--- a/summit/multiview_platform/multiview_classifiers/mumbo.py
+++ b/summit/multiview_platform/multiview_classifiers/mumbo.py
@@ -47,6 +47,13 @@ class Mumbo(BaseMultiviewClassifier, MumboClassifier):
"""
if base_estimator is None:
self.base_estimator = DecisionTreeClassifier()
+ elif type(base_estimator) is list:
+ if type(base_estimator[0]) is dict:
+ self.base_estimator = [self.set_base_estim_from_dict(estim) for estim in base_estimator]
+ elif isinstance(base_estimator[0], BaseEstimator):
+ self.base_estimator = base_estimator
+ else:
+ raise ValueError("base_estimator should ba a list of dict or a sklearn classifier list")
elif isinstance(base_estimator, dict):
self.base_estimator = self.set_base_estim_from_dict(base_estimator)
MumboClassifier.set_params(self, **params)
--
GitLab